{"schema_version":"onlylabs.public_analysis.v1","url":"https://onlylabs.fyi/analysis/amazon","json_url":"https://onlylabs.fyi/analysis/amazon/analysis.json","evidence_json_url":"https://onlylabs.fyi/analysis/amazon/evidence.json","generated_at":"2026-06-11T15:10:05.921Z","analysis":{"org_slug":"amazon","url":"https://onlylabs.fyi/analysis/amazon","json_url":"https://onlylabs.fyi/analysis/amazon/analysis.json","evidence_json_url":"https://onlylabs.fyi/analysis/amazon/evidence.json","dossier_url":"https://onlylabs.fyi/labs/amazon","org":{"slug":"amazon","name":"Amazon (Nova)","category":"frontier-lab","category_label":"Frontier lab","homepage_url":"https://nova.amazon.com/"},"title":"Amazon (Nova) analysis","summary":"Amazon is betting its AI identity on agentic AI as the organizing principle. The evidence pack shows a coordinated push: multiple amazon.science posts on agent design patterns, open-source agent frameworks spanning RL training, multi-agent evolution, and compliance verification, and product launches including Nova Act and the perception agent harness. Alongside this, Amazon is building the trust infrastructure…","markdown":"```json\n{\n  \"content\": \"## Thesis\\n\\nAmazon is betting its AI identity on agentic AI as the organizing principle. The evidence pack shows a coordinated push: multiple amazon.science posts on agent design patterns [P1](https://www.amazon.science/blog/bridging-intent-and-execution-in-agentic-systems)[P2](https://www.amazon.science/blog/real-world-grounding-in-agentic-ai)[P10](https://www.amazon.science/blog/designing-ai-agents-that-know-when-to-step-back)[P11](https://www.amazon.science/blog/how-agentic-ai-helps-heal-the-systems-we-cant-replace)[P16](https://www.amazon.science/blog/how-amazon-uses-agentic-ai-for-vulnerability-detection-at-global-scale), open-source agent frameworks spanning RL training, multi-agent evolution, and compliance verification [E5](https://github.com/amazon-science/reskill)[E12](https://github.com/amazon-science/EvoMAS)[E24](https://github.com/amazon-science/compagent)[E55](https://github.com/amazon-science/agentic-forking-path), and product launches including Nova Act [W4](https://www.aboutamazon.com/news/aws/how-amazon-builds-ai-agents) and the perception agent harness [W1](https://labs.amazon.science/blog/introducing-the-perception-agent-harness-annotation-and-verification-open-source). Alongside this, Amazon is building the trust infrastructure needed for agent deployment at scale — formal verification [P13](https://www.amazon.science/blog/formally-verified-aes-xts-the-first-aes-algorithm-to-join-s2n-bignum)[P19](https://www.amazon.science/blog/isabelle-hol-the-proof-assistant-behind-the-nitro-isolation-engine)[E32](https://www.amazon.science/blog/isabelle-hol-the-proof-assistant-behind-the-nitro-isolation-engine), post-quantum cryptography [P15](https://www.amazon.science/blog/verifying-and-optimizing-post-quantum-cryptography-at-amazon)[E42](https://www.amazon.science/blog/verifying-and-optimizing-post-quantum-cryptography-at-amazon), privacy-preserving training with cryptographic defenses [P21](https://www.amazon.science/blog/preserving-the-privacy-of-ai-training-data)[E30](https://www.amazon.science/blog/preserving-the-privacy-of-ai-training-data), and LLM catastrophic risk certification [P20](https://www.amazon.science/blog/how-catastrophic-is-your-llm)[E31](https://www.amazon.science/blog/how-catastrophic-is-your-llm). The Nova model family serves both as a standalone offering (Chronos-2 at 12.5M downloads [E1](https://huggingface.co/amazon/chronos-2)) and a customization platform (Nova Forge hyperparameter optimization [W2](https://aws.amazon.com/blogs/machine-learning/the-art-and-science-of-hyperparameter-optimization-on-amazon-nova-forge/), Nova for molecular-property prediction [P18](https://www.amazon.science/blog/customized-amazon-nova-models-improve-molecular-property-prediction-in-drug-discovery)[E35](https://www.amazon.science/blog/customized-amazon-nova-models-improve-molecular-property-prediction-in-drug-discovery)). Amazon leverages its operational DNA — supply chain optimization [P23](https://www.amazon.science/blog/how-mechanism-design-theory-helps-optimize-amazon-vendor-collaboration)[P24](https://www.amazon.science/blog/navigating-uncertainty-in-amazons-middle-mile-network)[E22](https://www.amazon.science/blog/navigating-uncertainty-in-amazons-middle-mile-network), datacenter network innovation [P5](https://arxiv.org/pdf/2604.15261)[E2](https://www.amazon.science/blog/how-flat-is-replacing-fat-in-aws-data-center-networks), and security at scale [P16](https://www.amazon.science/blog/how-amazon-uses-agentic-ai-for-vulnerability-detection-at-global-scale)[E41](https://www.amazon.science/blog/how-amazon-uses-agentic-ai-for-vulnerability-detection-at-global-scale) — as its differentiation. Critically, **this pack contains zero hiring signals**, making workforce strategy an information gap.\\n\\n## Signal desks\\n\\n### Hiring\\n\\nNo cited evidence in this pack. No job listings, career pages, or role announcements appear across the 28 pages, 60 events, or 4 web search results.\\n\\n### Forks\\n\\nNo cited evidence in this pack. All GitHub activity consists of first-party repos published by `amazon-science`; no forked upstream repositories are identified in the evidence [E5](https://github.com/amazon-science/reskill)[E7](https://github.com/amazon-science/dualkv-flash-attn-for-rl)[E12](https://github.com/amazon-science/EvoMAS)[E17](https://github.com/amazon-science/adaptive-layerwise-perturbation)[E19](https://github.com/amazon-science/temporal-reasoning-dataset)[E20](https://github.com/amazon-science/PROF-GRPO)[E21](https://github.com/amazon-science/hallucination-benchmark-trivialplus)[E23](https://github.com/amazon-science/RecArena)[E24](https://github.com/amazon-science/compagent)[E25](https://github.com/amazon-science/SWAN)[E28](https://github.com/amazon-science/rmir)[E33](https://github.com/amazon-science/expert-upcycling)[E34](https://github.com/amazon-science/CodeStruct)[E39](https://github.com/amazon-science/TransitionFlowMatching)[E51](https://github.com/amazon-science/storm-referring-multi-object-grounding)[E54](https://github.com/amazon-science/acclaim)[E55](https://github.com/amazon-science/agentic-forking-path)[E57](https://github.com/amazon-science/papercode-coordinating-spot-and-contracts)[E60](https://github.com/amazon-science/TSFM-Biases).\\n\\n### Releases\\n\\n- **Chronos-2**: Time-series forecasting foundation model, 119M params, Apache 2.0, 12.5M HuggingFace downloads, 317 likes — the highest-traction artifact in the pack [E1](https://huggingface.co/amazon/chronos-2).\\n- **P-EAGLE speculative decoding series**: Long-context models (gpt-oss-20b/120b, Qwen3-Coder-30B) targeting inference efficiency via speculative decoding, all Apache 2.0 [E13](https://huggingface.co/amazon/gpt-oss-20b-p-eagle-long-context)[E14](https://huggingface.co/amazon/Qwen3-Coder-30B-A3B-Instruct-P-EAGLE-long-context)[E15](https://huggingface.co/amazon/gpt-oss-120b-p-eagle-long-context)[E56](https://huggingface.co/amazon/gpt-oss-120b-p-eagle).\\n- **HQwen3 primed fine-tune batch**: At least 10 models released 2026-03-31 using GKA, GDN, Mamba2, and BMOJOF priming methods on Qwen3 8B/32B backbones across Instruct and Reasoner variants, Apache 2.0 [E38](https://huggingface.co/amazon/GKA-primed-HQwen3-32B-Instruct)[E40](https://huggingface.co/amazon/Mamba2-primed-HQwen3-8B-Instruct)[E43](https://huggingface.co/amazon/GKA-primed-HQwen3-8B-Reasoner)[E44](https://huggingface.co/amazon/GKA-primed-HQwen3-32B-Reasoner)[E45](https://huggingface.co/amazon/GDN-primed-HQwen3-32B-Instruct)[E47](https://huggingface.co/amazon/GKA-primed-HQwen3-8B-Instruct)[E48](https://huggingface.co/amazon/GDN-primed-HQwen3-8B-Instruct)[E49](https://huggingface.co/amazon/BMOJOF-primed-HQwen3-8B-Instruct)[E50](https://huggingface.co/amazon/GDN-primed-HQwen3-8B-Reasoner).\\n- **Agent infrastructure**: `reskill` — veRL extension for agent RL training with skill co-evolution [E5](https://github.com/amazon-science/reskill); `EvoMAS` — evolutionary multi-agent system generation, ICML 2026 [E12](https://github.com/amazon-science/EvoMAS); `CompAgent` — visual compliance verification agent [E24](https://github.com/amazon-science/compagent); `agentic-forking-path` [E55](https://github.com/amazon-science/agentic-forking-path).\\n- **Evaluation and dataset repos**: `temporal-reasoning-dataset` — multilingual temporal reasoning benchmark [E19](https://github.com/amazon-science/temporal-reasoning-dataset); `hallucination-benchmark-trivialplus` — ACL 2026 long-context hallucination detection benchmark [E21](https://github.com/amazon-science/hallucination-benchmark-trivialplus); `RMIR` — reasoning-intensive multimodal image retrieval benchmark [E28](https://github.com/amazon-science/rmir); `RecArena` [E23](https://github.com/amazon-science/RecArena).\\n- **Training infrastructure**: `dualkv-flash-attn-for-rl` — shared-prompt flash attention for efficient RL training [E7](https://github.com/amazon-science/dualkv-flash-attn-for-rl); `PROF-GRPO` [E20](https://github.com/amazon-science/PROF-GRPO); `expert-upcycling` (14 GitHub stars) [E33](https://github.com/amazon-science/expert-upcycling); `adaptive-layerwise-perturbation` [E17](https://github.com/amazon-science/adaptive-layerwise-perturbation).\\n- **Library releases with active cadence**: `concurry` v0.13.1/v0.13.2 [E10](https://github.com/amazon-science/concurry/releases/tag/v0.13.2)[E11](https://github.com/amazon-science/concurry/releases/tag/v0.13.1); `azcausal` v0.2.4.3/v0.2.5 [E29](https://github.com/amazon-science/azcausal/releases/tag/v0.2.5)[E52](https://github.com/amazon-science/azcausal/releases/tag/v0.2.4.3); `uniqsketch` v1.2.1 [E46](https://github.com/amazon-science/uniqsketch/releases/tag/v1.2.1).\\n- **Other notable repos**: `TransitionFlowMatching` — AISTATS 2026, image/video generation, 12 stars [E39](https://github.com/amazon-science/TransitionFlowMatching); `SWAN` — semantic watermarking, ACL 2026 [E25](https://github.com/amazon-science/SWAN); `CodeStruct` [E34](https://github.com/amazon-science/CodeStruct); `TSFM-Biases` — time-series foundation model bias analysis [E60](https://github.com/amazon-science/TSFM-Biases); `storm-referring-multi-object-grounding` [E51](https://github.com/amazon-science/storm-referring-multi-object-grounding); `acclaim` [E54](https://github.com/amazon-science/acclaim); `papercode-coordinating-spot-and-contracts` [E57](https://github.com/amazon-science/papercode-coordinating-spot-and-contracts).\\n\\n### Talking\\n\\n- **Agentic AI is the dominant narrative**: Posts cover bridging intent and execution in agentic systems [P1](https://www.amazon.science/blog/bridging-intent-and-execution-in-agentic-systems)[E4](https://www.amazon.science/blog/bridging-intent-and-execution-in-agentic-systems), four approaches to real-world grounding for AI agents [P2](https://www.amazon.science/blog/real-world-grounding-in-agentic-ai)[E3](https://www.amazon.science/blog/real-world-grounding-in-agentic-ai), UX design for human-AI coordination in agentic systems [P10](https://www.amazon.science/blog/designing-ai-agents-that-know-when-to-step-back), agentic AI for healing legacy systems that can't be replaced [P11](https://www.amazon.science/blog/how-agentic-ai-helps-heal-the-systems-we-cant-replace), RuleForge agentic vulnerability detection producing rules 336% faster [P16](https://www.amazon.science/blog/how-amazon-uses-agentic-ai-for-vulnerability-detection-at-global-scale)[E41](https://www.amazon.science/blog/how-amazon-uses-agentic-ai-for-vulnerability-detection-at-global-scale), the open-source perception agent harness [W1](https://labs.amazon.science/blog/introducing-the-perception-agent-harness-annotation-and-verification-open-source), and Amazon's overall agentic-AI approach with Nova Act training model capabilities, orchestration, and tool controls as one integrated system [W4](https://www.aboutamazon.com/news/aws/how-amazon-builds-ai-agents).\\n- **Trust, safety, and formal verification stack**: Amazon's responsible-AI pipeline embedding safety throughout the development lifecycle [P22](https://www.amazon.science/blog/building-trust-into-ai)[E27](https://www.amazon.science/blog/building-trust-into-ai); statistical framework for certifying LLM catastrophic failure likelihood in adversarial conversations [P20](https://www.amazon.science/blog/how-catastrophic-is-your-llm)[E31](https://www.amazon.science/blog/how-catastrophic-is-your-llm); reproducing training-data extraction attacks and cryptographic defenses that stop them [P21](https://www.amazon.science/blog/preserving-the-privacy-of-ai-training-data)[E30](https://www.amazon.science/blog/preserving-the-privacy-of-ai-training-data); formally verified AES-XTS as first AES algorithm in s2n-bignum [P13](https://www.amazon.science/blog/formally-verified-aes-xts-the-first-aes-algorithm-to-join-s2n-bignum)[E58](https://www.amazon.science/blog/formally-verified-aes-xts-the-first-aes-algorithm-to-join-s2n-bignum); verifying and optimizing post-quantum cryptography with automated reasoning [P15](https://www.amazon.science/blog/verifying-and-optimizing-post-quantum-cryptography-at-amazon)[E42](https://www.amazon.science/blog/verifying-and-optimizing-post-quantum-cryptography-at-amazon); Isabelle/HOL proof assistant enabling the world's first formally verified cloud hypervisor (Nitro Isolation Engine) [P19](https://www.amazon.science/blog/isabelle-hol-the-proof-assistant-behind-the-nitro-isolation-engine)[E32](https://www.amazon.science/blog/isabelle-hol-the-proof-assistant-behind-the-nitro-isolation-engine); academic collaboration delivering real-world security to customers [P6](https://www.amazon.science/news/how-academic-collaboration-delivers-real-world-security-to-amazon-customers).\\n- **Inference and training efficiency**: New scaling law connecting architectural choices to loss, identifying models with up to 47% throughput improvement at no accuracy loss [P26](https://www.amazon.science/blog/making-llms-faster-without-sacrificing-accuracy)[E16](https://www.amazon.science/blog/making-llms-faster-without-sacrificing-accuracy); thesis that intelligence isn't about parameter count but inference time — larger models become less insightful without reduced inference time [P8](https://www.amazon.science/blog/intelligence-isnt-about-parameter-count-its-about-time)[E37](https://www.amazon.science/blog/intelligence-isnt-about-parameter-count-its-about-time); LoRA target module selection ablation study on accuracy-efficiency trade-offs [P12](https://www.amazon.science/blog/optimizing-lora-target-module-selection-for-efficient-fine-tuning)[E59](https://www.amazon.science/blog/optimizing-lora-target-module-selection-for-efficient-fine-tuning); Promptimus automated prompt-engineering framework for improving prompts without manual work [P25](https://www.amazon.science/blog/promptimus-improving-already-good-llm-prompts-with-zero-manual-engineering)[E18](https://www.amazon.science/blog/promptimus-improving-already-good-llm-prompts-with-zero-manual-engineering); training LLMs to generate diverse accurate reasoning paths using global forking tokens [P27](https://www.amazon.science/blog/diverse-reasoning-traces-teach-llms-to-make-better-decisions)[E9](https://www.amazon.science/blog/diverse-reasoning-traces-teach-llms-to-make-better-decisions).\\n- **Operational optimization**: Mechanism design theory applied to Amazon-vendor supply chain collaboration without disclosing private information [P23](https://www.amazon.science/blog/how-mechanism-design-theory-helps-optimize-amazon-vendor-collaboration)[E26](https://www.amazon.science/blog/how-mechanism-design-theory-helps-optimize-amazon-vendor-collaboration); new tools for optimizing middle-mile delivery networks under uncertainty [P24](https://www.amazon.science/blog/navigating-uncertainty-in-amazons-middle-mile-network)[E22](https://www.amazon.science/blog/navigating-uncertainty-in-amazons-middle-mile-network); RNG flat datacenter networks using quasi-random graphs and ShuffleBox optical devices, now default for most AWS workloads, up to 45% cheaper than fat trees [P5](https://arxiv.org/pdf/2604.15261)[E2](https://www.amazon.science/blog/how-flat-is-replacing-fat-in-aws-data-center-networks); 12-year-old forecasting paper still proving durable [P7](https://www.amazon.science/blog/why-a-12-year-old-forecasting-paper-has-stood-the-test-of-time).\\n- **Domain applications**: Customized Amazon Nova models unifying molecular-property prediction in drug discovery, serving as reasoning partner for medical chemists [P18](https://www.amazon.science/blog/customized-amazon-nova-models-improve-molecular-property-prediction-in-drug-discovery)[E35](https://www.amazon.science/blog/customized-amazon-nova-models-improve-molecular-property-prediction-in-drug-discovery); AWS–Johns Hopkins antibody developability benchmark with diverse public antibody datasets for AI-guided antibody design [P17](https://www.amazon.science/news/aws-gray-lab-johns-hopkins-announce-groundbreaking-database-for-ai-ml-antibody-design)[E36](https://www.amazon.science/news/aws-gray-lab-johns-hopkins-announce-groundbreaking-database-for-ai-ml-antibody-design); LLM-based TTS improvements via LoRA, data augmentation, and chain-of-thought reasoning for accent-free polyglot output [P14](https://www.amazon.science/blog/improving-quality-and-robustness-in-llm-based-text-to-speech-systems)[E53](https://www.amazon.science/blog/improving-quality-and-robustness-in-llm-based-text-to-speech-systems); AI changing the nature of mathematical research [P9](https://www.amazon.science/blog/how-ai-is-changing-the-nature-of-mathematical-research).\\n- **Data and evaluation**: Ground truth framed as a process, not a dataset — challenges in auto-fact-checking long AI-generated research reports [E6](https://www.amazon.science/blog/ground-truth-is-a-process-not-a-dataset); Nova Sonic Test Harness for evaluating voice agents at scale with audio-hallucination detection and LLM-as-judge [W3](https://aws.amazon.com/blogs/machine-learning/evaluate-your-amazon-nova-sonic-voice-agent-at-scale-no-microphone-required/); hyperparameter optimization on Amazon Nova Forge covering data mixing, learning rate, checkpoint selection [W2](https://aws.amazon.com/blogs/machine-learning/the-art-and-science-of-hyperparameter-optimization-on-amazon-nova-forge/); Amazon Research Awards funding recipients across 49 universities in 11 countries with access to Amazon public datasets and AWS AI/ML services [P28](https://www.amazon.science/research-awards/latest-news/fall-2025-amazon-research-awards-recipients-announced)[E8](https://www.amazon.science/research-awards/latest-news/fall-2025-amazon-research-awards-recipients-announced).\\n\\n## Shipping\\n\\nAmazon ships across four lanes in this evidence window:\\n\\n1. **Models**: Chronos-2 dominates with 12.5M downloads [E1](https://huggingface.co/amazon/chronos-2); P-EAGLE speculative decoding series across GPT-OSS and Qwen3-Coder backbones [E13](https://huggingface.co/amazon/gpt-oss-20b-p-eagle-long-context)[E14](https://huggingface.co/amazon/Qwen3-Coder-30B-A3B-Instruct-P-EAGLE-long-context)[E15](https://huggingface.co/amazon/gpt-oss-120b-p-eagle-long-context)[E56](https://huggingface.co/amazon/gpt-oss-120b-p-eagle); a large batch of primed HQwen3 fine-tunes using GKA, GDN, Mamba2, and BMOJOF methods [E38](https://huggingface.co/amazon/GKA-primed-HQwen3-32B-Instruct)[E40](https://huggingface.co/amazon/Mamba2-primed-HQwen3-8B-Instruct)[E43](https://huggingface.co/amazon/GKA-primed-HQwen3-8B-Reasoner)[E44](https://huggingface.co/amazon/GKA-primed-HQwen3-32B-Reasoner)[E45](https://huggingface.co/amazon/GDN-primed-HQwen3-32B-Instruct)[E47](https://huggingface.co/amazon/GKA-primed-HQwen3-8B-Instruct)[E48](https://huggingface.co/amazon/GDN-primed-HQwen3-8B-Instruct)[E49](https://huggingface.co/amazon/BMOJOF-primed-HQwen3-8B-Instruct)[E50](https://huggingface.co/amazon/GDN-primed-HQwen3-8B-Reasoner).\\n2. **Agent frameworks**: `reskill` for agent RL with skill co-evolution [E5](https://github.com/amazon-science/reskill); `EvoMAS` for evolutionary multi-agent systems [E12](https://github.com/amazon-science/EvoMAS); `CompAgent` for visual compliance [E24](https://github.com/amazon-science/compagent); perception agent harness with annotation and verification primitives [W1](https://labs.amazon.science/blog/introducing-the-perception-agent-harness-annotation-and-verification-open-source); Nova Act as an integrated agent-building service [W4](https://www.aboutamazon.com/news/aws/how-amazon-builds-ai-agents).\\n3. **Evaluation infrastructure**: Hallucination detection benchmarks [E21](https://github.com/amazon-science/hallucination-benchmark-trivialplus), temporal reasoning datasets [E19](https://github.com/amazon-science/temporal-reasoning-dataset), multimodal retrieval benchmarks [E28](https://github.com/amazon-science/rmir), Nova Sonic test harness with audio-hallucination detection [W3](https://aws.amazon.com/blogs/machine-learning/evaluate-your-amazon-nova-sonic-voice-agent-at-scale-no-microphone-required/), Antibody Developability Benchmark [E36](https://www.amazon.science/news/aws-gray-lab-johns-hopkins-announce-groundbreaking-database-for-ai-ml-antibody-design).\\n4. **Core infrastructure**: RNG flat datacenter networks now default for most AWS workloads, up to 45% cheaper [P5](https://arxiv.org/pdf/2604.15261)[E2](https://www.amazon.science/blog/how-flat-is-replacing-fat-in-aws-data-center-networks); formally verified AES-XTS in s2n-bignum [P13](https://www.amazon.science/blog/formally-verified-aes-xts-the-first-aes-algorithm-to-join-s2n-bignum)[E58](https://www.amazon.science/blog/formally-verified-aes-xts-the-first-aes-algorithm-to-join-s2n-bignum); `concurry`, `azcausal`, and `uniqsketch` library releases on active cadences [E10](https://github.com/amazon-science/concurry/releases/tag/v0.13.2)[E11](https://github.com/amazon-science/concurry/releases/tag/v0.13.1)[E29](https://github.com/amazon-science/azcausal/releases/tag/v0.2.5)[E46](https://github.com/amazon-science/uniqsketch/releases/tag/v1.2.1)[E52](https://github.com/amazon-science/azcausal/releases/tag/v0.2.4.3).\\n\\n## Research themes\\n\\n- **Agentic AI systems**: Design patterns for intent-execution bridging [P1](https://www.amazon.science/blog/bridging-intent-and-execution-in-agentic-systems)[E4](https://www.amazon.science/blog/bridging-intent-and-execution-in-agentic-systems), real-world grounding approaches [P2](https://www.amazon.science/blog/real-world-grounding-in-agentic-ai)[E3](https://www.amazon.science/blog/real-world-grounding-in-agentic-ai), human-AI coordination UX [P10](https://www.amazon.science/blog/designing-ai-agents-that-know-when-to-step-back), legacy-system integration [P11](https://www.amazon.science/blog/how-agentic-ai-helps-heal-the-systems-we-cant-replace), multi-agent evolutionary systems [E12](https://github.com/amazon-science/EvoMAS), and agentic forking-path architectures [E55](https://github.com/amazon-science/agentic-forking-path).\\n- **Trustworthy AI stack**: Formal verification with Isabelle/HOL for cloud hypervisors [P19](https://www.amazon.science/blog/isabelle-hol-the-proof-assistant-behind-the-nitro-isolation-engine)[E32](https://www.amazon.science/blog/isabelle-hol-the-proof-assistant-behind-the-nitro-isolation-engine); verified AES-XTS and post-quantum cryptography [P13](https://www.amazon.science/blog/formally-verified-aes-xts-the-first-aes-algorithm-to-join-s2n-bignum)[P15](https://www.amazon.science/blog/verifying-and-optimizing-post-quantum-cryptography-at-amazon)[E42](https://www.amazon.science/blog/verifying-and-optimizing-post-quantum-cryptography-at-amazon)[E58](https://www.amazon.science/blog/formally-verified-aes-xts-the-first-aes-algorithm-to-join-s2n-bignum); responsible-AI pipeline development [P22](https://www.amazon.science/blog/building-trust-into-ai)[E27](https://www.amazon.science/blog/building-trust-into-ai); LLM catastrophic risk certification through statistical frameworks [P20](https://www.amazon.science/blog/how-catastrophic-is-your-llm)[E31](https://www.amazon.science/blog/how-catastrophic-is-your-llm); cryptographic defenses against training-data extraction [P21](https://www.amazon.science/blog/preserving-the-privacy-of-ai-training-data)[E30](https://www.amazon.science/blog/preserving-the-privacy-of-ai-training-data); semantic watermarking [E25](https://github.com/amazon-science/SWAN).\\n- **Efficient training and inference**: Scaling laws linking architecture to inference throughput [P26](https://www.amazon.science/blog/making-llms-faster-without-sacrificing-accuracy)[E16](https://www.amazon.science/blog/making-llms-faster-without-sacrificing-accuracy); speculative decoding via P-EAGLE [E13](https://huggingface.co/amazon/gpt-oss-20b-p-eagle-long-context)[E14](https://huggingface.co/amazon/Qwen3-Coder-30B-A3B-Instruct-P-EAGLE-long-context)[E15](https://huggingface.co/amazon/gpt-oss-120b-p-eagle-long-context); LoRA target module optimization [P12](https://www.amazon.science/blog/optimizing-lora-target-module-selection-for-efficient-fine-tuning)[E59](https://www.amazon.science/blog/optimizing-lora-target-module-selection-for-efficient-fine-tuning); RL training efficiency with DualKV flash attention [E7](https://github.com/amazon-science/dualkv-flash-attn-for-rl); expert upcycling [E33](https://github.com/amazon-science/expert-upcycling); inference-time intelligence thesis [P8](https://www.amazon.science/blog/intelligence-isnt-about-parameter-count-its-about-time)[E37](https://www.amazon.science/blog/intelligence-isnt-about-parameter-count-its-about-time); diverse reasoning trace training [P27](https://www.amazon.science/blog/diverse-reasoning-traces-teach-llms-to-make-better-decisions)[E9](https://www.amazon.science/blog/diverse-reasoning-traces-teach-llms-to-make-better-decisions).\\n- **Domain-specific AI**: Time-series forecasting via Chronos-2 [E1](https://huggingface.co/amazon/chronos-2); drug discovery with customized Nova models [P18](https://www.amazon.science/blog/customized-amazon-nova-models-improve-molecular-property-prediction-in-drug-discovery)[E35](https://www.amazon.science/blog/customized-amazon-nova-models-improve-molecular-property-prediction-in-drug-discovery); antibody design benchmarking [P17](https://www.amazon.science/news/aws-gray-lab-johns-hopkins-announce-groundbreaking-database-for-ai-ml-antibody-design)[E36](https://www.amazon.science/news/aws-gray-lab-johns-hopkins-announce-groundbreaking-database-for-ai-ml-antibody-design); LLM-based text-to-speech quality and robustness [P14](https://www.amazon.science/blog/improving-quality-and-robustness-in-llm-based-text-to-speech-systems)[E53](https://www.amazon.science/blog/improving-quality-and-robustness-in-llm-based-text-to-speech-systems); mechanism design for supply chain [P23](https://www.amazon.science/blog/how-mechanism-design-theory-helps-optimize-amazon-vendor-collaboration)[E26](https://www.amazon.science/blog/how-mechanism-design-theory-helps-optimize-amazon-vendor-collaboration).\\n- **Systems and optimization science**: Quasi-random graph datacenter networks [P5](https://arxiv.org/pdf/2604.15261)[E2](https://www.amazon.science/blog/how-flat-is-replacing-fat-in-aws-data-center-networks); middle-mile logistics under uncertainty [P24](https://www.amazon.science/blog/navigating-uncertainty-in-amazons-middle-mile-network)[E22](https://www.amazon.science/blog/navigating-uncertainty-in-amazons-middle-mile-network); forecasting methodology [P7](https://www.amazon.science/blog/why-a-12-year-old-forecasting-paper-has-stood-the-test-of-time); automated prompt engineering [P25](https://www.amazon.science/blog/promptimus-improving-already-good-llm-prompts-with-zero-manual-engineering)[E18](https://www.amazon.science/blog/promptimus-improving-already-good-llm-prompts-with-zero-manual-engineering); causal inference libraries [E29](https://github.com/amazon-science/azcausal/releases/tag/v0.2.5)[E52](https://github.com/amazon-science/azcausal/releases/tag/v0.2.4.3).\\n\\n## Hiring & scaling\\n\\nNo hiring signals appear in this evidence pack. The pattern of 30+ open-source repos from `amazon-science` and sustained blog output from `amazon.science` suggests an active, publishing research organization [E5](https://github.com/amazon-science/reskill)[E7](https://github.com/amazon-science/dualkv-flash-attn-for-rl)[E12](https://github.com/amazon-science/EvoMAS)[E17](https://github.com/amazon-science/adaptive-layerwise-perturbation)[E19](https://github.com/amazon-science/temporal-reasoning-dataset)[E20](https://github.com/amazon-science/PROF-GRPO)[E21](https://github.com/amazon-science/hallucination-benchmark-trivialplus)[E23](https://github.com/amazon-science/RecArena)[E24](https://github.com/amazon-science/compagent)[E25](https://github.com/amazon-science/SWAN)[E28](https://github.com/amazon-science/rmir)[E33](https://github.com/amazon-science/expert-upcycling)[E34](https://github.com/amazon-science/CodeStruct)[E39](https://github.com/amazon-science/TransitionFlowMatching)[E51](https://github.com/amazon-science/storm-referring-multi-object-grounding)[E54](https://github.com/amazon-science/acclaim)[E55](https://github.com/amazon-science/agentic-forking-path)[E57](https://github.com/amazon-science/papercode-coordinating-spot-and-contracts)[E60](https://github.com/amazon-science/TSFM-Biases), but roles, locations, team sizes, headcount growth, and geographic hubs cannot be estimated from the supplied evidence. The Amazon Research Awards program engages 49 universities across 11 countries [P28](https://www.amazon.science/research-awards/latest-news/fall-2025-amazon-research-awards-recipients-announced)[E8](https://www.amazon.science/research-awards/latest-news/fall-2025-amazon-research-awards-recipients-announced), which may serve as an academic pipeline, but no conversion data into direct hiring is available. This is a notable gap for operators tracking Amazon's AI workforce buildout.\\n\\n## Data-business implications\\n\\n- **Data demand**: Chronos-2's 12.5M downloads signal strong enterprise appetite for time-series foundation models — an opportunity for curated forecasting dataset products [E1](https://huggingface.co/amazon/chronos-2). The Antibody Developability Benchmark [E36](https://www.amazon.science/news/aws-gray-lab-johns-hopkins-announce-groundbreaking-database-for-ai-ml-antibody-design) and hallucination-benchmark-trivialplus [E21](https://github.com/amazon-science/hallucination-benchmark-trivialplus) create new structured evaluation datasets; the temporal-reasoning-dataset spans multilingual benchmarks [E19](https://github.com/amazon-science/temporal-reasoning-dataset); RMIR extends to multimodal retrieval evaluation [E28](https://github.com/amazon-science/rmir). Amazon Research Awards grant 49 universities access to Amazon public datasets, expanding the data ecosystem [E8](https://www.amazon.science/research-awards/latest-news/fall-2025-amazon-research-awards-recipients-announced). Nova Forge's data mixing capability blends customer training data with curated datasets to prevent catastrophic forgetting during domain customization [W2](https://aws.amazon.com/blogs/machine-learning/the-art-and-science-of-hyperparameter-optimization-on-amazon-nova-forge/).\\n- **Evals and quality**: Nova Sonic Test Harness introduces audio-hallucination detection and LLM-as-judge evaluation at scale for voice agents [W3](https://aws.amazon.com/blogs/machine-learning/evaluate-your-amazon-nova-sonic-voice-agent-at-scale-no-microphone-required/). The \\\"ground truth is a process\\\" framing signals evolving eval methodologies beyond static benchmarks [E6](https://www.amazon.science/blog/ground-truth-is-a-process-not-a-dataset). The hallucination detection benchmark explicitly targets long-context RAG-based evaluation [E21](https://github.com/amazon-science/hallucination-benchmark-trivialplus). These create tooling opportunities for automated quality pipelines.\\n- **Infrastructure**: RNG flat networks — now default for most AWS workloads and up to 45% cheaper than fat trees — represent a datacenter topology shift with implications for training and inference cluster design [P5](https://arxiv.org/pdf/2604.15261)[E2](https://www.amazon.science/blog/how-flat-is-replacing-fat-in-aws-data-center-networks). The scaling law connecting architecture to inference throughput (47% improvement) [P26](https://www.amazon.science/blog/making-llms-faster-without-sacrificing-accuracy)[E16](https://www.amazon.science/blog/making-llms-faster-without-sacrificing-accuracy) and DualKV flash attention for RL training with large rollouts [E7](https://github.com/amazon-science/dualkv-flash-attn-for-rl) point to specialized infrastructure needs for RL-based agent training at scale. `reskill` extends veRL for agent RL with skill co-evolution [E5](https://github.com/amazon-science/reskill).\\n- **Tooling**: Nova Forge's hyperparameter optimization framework addresses learning rate, data mixing ratio, checkpoint selection, and training technique interactions [W2](https://aws.amazon.com/blogs/machine-learning/the-art-and-science-of-hyperparameter-optimization-on-amazon-nova-forge/). Promptimus provides automated prompt engineering by targeting specific failure points [E18](https://www.amazon.science/blog/promptimus-improving-already-good-llm-prompts-with-zero-manual-engineering). `concurry` [E10](https://github.com/amazon-science/concurry/releases/tag/v0.13.2)[E11](https://github.com/amazon-science/concurry/releases/tag/v0.13.1) and `azcausal` [E29](https://github.com/amazon-science/azcausal/releases/tag/v0.2.5)[E52](https://github.com/amazon-science/azcausal/releases/tag/v0.2.4.3) are utility libraries with active release cadences suitable for integration into data and ML platform toolchains.\\n- **Safety and deployment**: The responsible-AI pipeline [P22](https://www.amazon.science/blog/building-trust-into-ai)[E27](https://www.amazon.science/blog/building-trust-into-ai), LLM catastrophic risk certification framework [P20](https://www.amazon.science/blog/how-catastrophic-is-your-llm)[E31](https://www.amazon.science/blog/how-catastrophic-is-your-llm), privacy-preserving training with cryptographic defenses [P21](https://www.amazon.science/blog/preserving-the-privacy-of-ai-training-data)[E30](https://www.amazon.science/blog/preserving-the-privacy-of-ai-training-data), and CompAgent for visual compliance verification [E24](https://github.com/amazon-science/compagent) create safety tooling and governance infrastructure opportunities. Nova Act's reliability-first design, training model capabilities and orchestration together as one integrated system [W4](https://www.aboutamazon.com/news/aws/how-amazon-builds-ai-agents), and the perception agent harness [W1](https://labs.amazon.science/blog/introducing-the-perception-agent-harness-annotation-and-verification-open-source) position agentic AI as a product surface requiring new monitoring, evaluation, and guardrail infrastructure.\\n- **Deployment optimization**: P-EAGLE speculative decoding models across parameter scales (20B to 120B) [E13](https://huggingface.co/amazon/gpt-oss-20b-p-eagle-long-context)[E14](https://huggingface.co/amazon/Qwen3-Coder-30B-A3B-Instruct-P-EAGLE-long-context)[E15](https://huggingface.co/amazon/gpt-oss-120b-p-eagle-long-context)[E56](https://huggingface.co/amazon/gpt-oss-120b-p-eagle) and the inference-time intelligence thesis [E37](https://www.amazon.science/blog/intelligence-isnt-about-parameter-count-its-about-time) indicate deployment optimization around latency-sensitive agent workloads. The HQwen3 primed-series using GKA, GDN, Mamba2, and BMOJOF priming methods [E38](https://huggingface.co/amazon/GKA-primed-HQwen3-32B-Instruct)[E40](https://huggingface.co/amazon/Mamba2-primed-HQwen3-8B-Instruct)[E43](https://huggingface.co/amazon/GKA-primed-HQwen3-8B-Reasoner)[E44](https://huggingface.co/amazon/GKA-primed-HQwen3-32B-Reasoner)[E45](https://huggingface.co/amazon/GDN-primed-HQwen3-32B-Instruct)[E47](https://huggingface.co/amazon/GKA-primed-HQwen3-8B-Instruct)[E48](https://huggingface.co/amazon/GDN-primed-HQwen3-8B-Instruct)[E49](https://huggingface.co/amazon/BMOJOF-primed-HQwen3-8B-Instruct)[E50](https://huggingface.co/amazon/GDN-primed-HQwen3-8B-Reasoner) reflects systematic exploration of efficient deployment architectures that could inform serving infrastructure decisions.\\n\\n## Traction highlights\\n\\n- **Chronos-2**: 12.5M HuggingFace downloads, 317 likes — the standout traction artifact [E1](https://huggingface.co/amazon/chronos-2).\\n- **GKA-primed-HQwen3-32B-Instruct**: 61,931 downloads [E38](https://huggingface.co/amazon/GKA-primed-HQwen3-32B-Instruct).\\n- **GKA-primed-HQwen3-8B-Reasoner**: 3,941 downloads [E43](https://huggingface.co/amazon/GKA-primed-HQwen3-8B-Reasoner).\\n- **GKA-primed-HQwen3-8B-Instruct**: 3,241 downloads [E47](https://huggingface.co/amazon/GKA-primed-HQwen3-8B-Instruct).\\n- **GDN-primed-HQwen3-8B-Instruct**: 1,339 downloads [E48](https://huggingface.co/amazon/GDN-primed-HQwen3-8B-Instruct).\\n- **expert-upcycling**: 14 GitHub stars [E33](https://github.com/amazon-science/expert-upcycling).\\n- **TransitionFlowMatching**: 12 GitHub stars [E39](https://github.com/amazon-science/TransitionFlowMatching).\\n- **reskill**: 5 GitHub stars [E5](https://github.com/amazon-science/reskill).\\n- **HN engagement modest**: RNG flat networks post drew 4 points/2 comments [E2](https://www.amazon.science/blog/how-flat-is-replacing-fat-in-aws-data-center-networks); inference-time intelligence post drew 3 points/0 comments [E37](https://www.amazon.science/blog/intelligence-isnt-about-parameter-count-its-about-time).\\n- **Most newer repos have low star counts** (1–4 stars), suggesting early-stage research artifacts rather than production-adopted tooling [E7](https://github.com/amazon-science/dualkv-flash-attn-for-rl)[E12](https://github.com/amazon-science/EvoMAS)[E17](https://github.com/amazon-science/adaptive-layerwise-perturbation)[E19](https://github.com/amazon-science/temporal-reasoning-dataset)[E20](https://github.com/amazon-science/PROF-GRPO)[E21](https://github.com/amazon-science/hallucination-benchmark-trivialplus)[E23](https://github.com/amazon-science/RecArena)[E24](https://github.com/amazon-science/compagent)[E28](https://github.com/amazon-science/rmir)[E34](https://github.com/amazon-science/CodeStruct)[E51](https://github.com/amazon-science/storm-referring-multi-object-grounding)[E54](https://github.com/amazon-science/acclaim)[E55](https://github.com/amazon-science/agentic-forking-path)[E60](https://github.com/amazon-science/TSFM-Biases).\\n\\n## Sources\\n\\nP1, P2, P5, P6, P7, P8, P9, P10, P11, P12, P13, P14, P15, P16, P17, P18, P19, P20, P21, P22, P23, P24, P25, P26, P27, P28, E1, E2, E3, E4, E5, E6, E7, E8, E9, E10, E11, E12, E13, E14, E15, E16, E17, E18, E19, E20, E21, E22, E23, E24, E25, E26, E27, E28, E29, E30, E31, E32, E33, E34, E35, E36, E37, E38, E39, E40, E41, E42, E43, E44, E45, E46, E47, E48, E49, E50, E51, E52, E53, E54, E55, E56, E57, E58, E59, E60, W1, W2, W3, W4\",\n  \"cites\": [\n    \"P1\", \"P2\", \"P5\", \"P6\", \"P7\", \"P8\", \"P9\", \"P10\", \"P11\", \"P12\", \"P13\", \"P14\", \"P15\", \"P16\", \"P17\", \"P18\", \"P19\", \"P20\", \"P21\", \"P22\", \"P23\", \"P24\", \"P25\", \"P26\", \"P27\", \"P28\",\n    \"E1\", \"E2\", \"E3\", \"E4\", \"E5\", \"E6\", \"E7\", \"E8\", \"E9\", \"E10\", \"E11\", \"E12\", \"E13\", \"E14\", \"E15\", \"E16\", \"E17\", \"E18\", \"E19\", \"E20\", \"E21\", \"E22\", \"E23\", \"E24\", \"E25\", \"E26\", \"E27\", \"E28\", \"E29\", \"E30\", \"E31\", \"E32\", \"E33\", \"E34\", \"E35\", \"E36\", \"E37\", \"E38\", \"E39\", \"E40\", \"E41\", \"E42\", \"E43\", \"E44\", \"E45\", \"E46\", \"E47\", \"E48\", \"E49\", \"E50\", \"E51\", \"E52\", \"E53\", \"E54\", \"E55\", \"E56\", \"E57\", \"E58\", \"E59\", \"E60\",\n    \"W1\", \"W2\", \"W3\", \"W4\"\n  ]\n}\n```","generated_at":"2026-06-10T08:03:29.623+00:00","citations":[{"url":"https://www.amazon.science/blog/bridging-intent-and-execution-in-agentic-systems","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/real-world-grounding-in-agentic-ai","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/designing-ai-agents-that-know-when-to-step-back","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/how-agentic-ai-helps-heal-the-systems-we-cant-replace","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/how-amazon-uses-agentic-ai-for-vulnerability-detection-at-global-scale","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://github.com/amazon-science/reskill","path":null,"label":"amazon-science/reskill","type":"external"},{"url":"https://github.com/amazon-science/EvoMAS","path":null,"label":"amazon-science/EvoMAS","type":"external"},{"url":"https://github.com/amazon-science/compagent","path":null,"label":"amazon-science/compagent","type":"external"},{"url":"https://github.com/amazon-science/agentic-forking-path","path":null,"label":"amazon-science/agentic-forking-path","type":"external"},{"url":"https://www.aboutamazon.com/news/aws/how-amazon-builds-ai-agents","path":null,"label":"aboutamazon.com/news","type":"external"},{"url":"https://labs.amazon.science/blog/introducing-the-perception-agent-harness-annotation-and-verification-open-source","path":null,"label":"labs.amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/formally-verified-aes-xts-the-first-aes-algorithm-to-join-s2n-bignum","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/isabelle-hol-the-proof-assistant-behind-the-nitro-isolation-engine","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/isabelle-hol-the-proof-assistant-behind-the-nitro-isolation-engine","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/verifying-and-optimizing-post-quantum-cryptography-at-amazon","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/verifying-and-optimizing-post-quantum-cryptography-at-amazon","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/preserving-the-privacy-of-ai-training-data","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/preserving-the-privacy-of-ai-training-data","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/how-catastrophic-is-your-llm","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/how-catastrophic-is-your-llm","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://huggingface.co/amazon/chronos-2","path":null,"label":"amazon/chronos-2","type":"external"},{"url":"https://aws.amazon.com/blogs/machine-learning/the-art-and-science-of-hyperparameter-optimization-on-amazon-nova-forge/","path":null,"label":"aws.amazon.com/blogs","type":"external"},{"url":"https://www.amazon.science/blog/customized-amazon-nova-models-improve-molecular-property-prediction-in-drug-discovery","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/customized-amazon-nova-models-improve-molecular-property-prediction-in-drug-discovery","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/how-mechanism-design-theory-helps-optimize-amazon-vendor-collaboration","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/navigating-uncertainty-in-amazons-middle-mile-network","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/navigating-uncertainty-in-amazons-middle-mile-network","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://arxiv.org/pdf/2604.15261","path":null,"label":"arxiv.org/pdf","type":"external"},{"url":"https://www.amazon.science/blog/how-flat-is-replacing-fat-in-aws-data-center-networks","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/how-amazon-uses-agentic-ai-for-vulnerability-detection-at-global-scale","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://github.com/amazon-science/dualkv-flash-attn-for-rl","path":null,"label":"amazon-science/dualkv-flash-attn-for-rl","type":"external"},{"url":"https://github.com/amazon-science/adaptive-layerwise-perturbation","path":null,"label":"amazon-science/adaptive-layerwise-perturbation","type":"external"},{"url":"https://github.com/amazon-science/temporal-reasoning-dataset","path":null,"label":"amazon-science/temporal-reasoning-dataset","type":"external"},{"url":"https://github.com/amazon-science/PROF-GRPO","path":null,"label":"amazon-science/PROF-GRPO","type":"external"},{"url":"https://github.com/amazon-science/hallucination-benchmark-trivialplus","path":null,"label":"amazon-science/hallucination-benchmark-trivialplus","type":"external"},{"url":"https://github.com/amazon-science/RecArena","path":null,"label":"amazon-science/RecArena","type":"external"},{"url":"https://github.com/amazon-science/SWAN","path":null,"label":"amazon-science/SWAN","type":"external"},{"url":"https://github.com/amazon-science/rmir","path":null,"label":"amazon-science/rmir","type":"external"},{"url":"https://github.com/amazon-science/expert-upcycling","path":null,"label":"amazon-science/expert-upcycling","type":"external"},{"url":"https://github.com/amazon-science/CodeStruct","path":null,"label":"amazon-science/CodeStruct","type":"external"},{"url":"https://github.com/amazon-science/TransitionFlowMatching","path":null,"label":"amazon-science/TransitionFlowMatching","type":"external"},{"url":"https://github.com/amazon-science/storm-referring-multi-object-grounding","path":null,"label":"amazon-science/storm-referring-multi-object-grounding","type":"external"},{"url":"https://github.com/amazon-science/acclaim","path":null,"label":"amazon-science/acclaim","type":"external"},{"url":"https://github.com/amazon-science/papercode-coordinating-spot-and-contracts","path":null,"label":"amazon-science/papercode-coordinating-spot-and-contracts","type":"external"},{"url":"https://github.com/amazon-science/TSFM-Biases","path":null,"label":"amazon-science/TSFM-Biases","type":"external"},{"url":"https://huggingface.co/amazon/gpt-oss-20b-p-eagle-long-context","path":null,"label":"amazon/gpt-oss-20b-p-eagle-long-context","type":"external"},{"url":"https://huggingface.co/amazon/Qwen3-Coder-30B-A3B-Instruct-P-EAGLE-long-context","path":null,"label":"amazon/Qwen3-Coder-30B-A3B-Instruct-P-EAGLE-long-context","type":"external"},{"url":"https://huggingface.co/amazon/gpt-oss-120b-p-eagle-long-context","path":null,"label":"amazon/gpt-oss-120b-p-eagle-long-context","type":"external"},{"url":"https://huggingface.co/amazon/gpt-oss-120b-p-eagle","path":null,"label":"amazon/gpt-oss-120b-p-eagle","type":"external"},{"url":"https://huggingface.co/amazon/GKA-primed-HQwen3-32B-Instruct","path":null,"label":"amazon/GKA-primed-HQwen3-32B-Instruct","type":"external"},{"url":"https://huggingface.co/amazon/Mamba2-primed-HQwen3-8B-Instruct","path":null,"label":"amazon/Mamba2-primed-HQwen3-8B-Instruct","type":"external"},{"url":"https://huggingface.co/amazon/GKA-primed-HQwen3-8B-Reasoner","path":null,"label":"amazon/GKA-primed-HQwen3-8B-Reasoner","type":"external"},{"url":"https://huggingface.co/amazon/GKA-primed-HQwen3-32B-Reasoner","path":null,"label":"amazon/GKA-primed-HQwen3-32B-Reasoner","type":"external"},{"url":"https://huggingface.co/amazon/GDN-primed-HQwen3-32B-Instruct","path":null,"label":"amazon/GDN-primed-HQwen3-32B-Instruct","type":"external"},{"url":"https://huggingface.co/amazon/GKA-primed-HQwen3-8B-Instruct","path":null,"label":"amazon/GKA-primed-HQwen3-8B-Instruct","type":"external"},{"url":"https://huggingface.co/amazon/GDN-primed-HQwen3-8B-Instruct","path":null,"label":"amazon/GDN-primed-HQwen3-8B-Instruct","type":"external"},{"url":"https://huggingface.co/amazon/BMOJOF-primed-HQwen3-8B-Instruct","path":null,"label":"amazon/BMOJOF-primed-HQwen3-8B-Instruct","type":"external"},{"url":"https://huggingface.co/amazon/GDN-primed-HQwen3-8B-Reasoner","path":null,"label":"amazon/GDN-primed-HQwen3-8B-Reasoner","type":"external"},{"url":"https://github.com/amazon-science/concurry/releases/tag/v0.13.2","path":null,"label":"amazon-science/concurry","type":"external"},{"url":"https://github.com/amazon-science/concurry/releases/tag/v0.13.1","path":null,"label":"amazon-science/concurry","type":"external"},{"url":"https://github.com/amazon-science/azcausal/releases/tag/v0.2.5","path":null,"label":"amazon-science/azcausal","type":"external"},{"url":"https://github.com/amazon-science/azcausal/releases/tag/v0.2.4.3","path":null,"label":"amazon-science/azcausal","type":"external"},{"url":"https://github.com/amazon-science/uniqsketch/releases/tag/v1.2.1","path":null,"label":"amazon-science/uniqsketch","type":"external"},{"url":"https://www.amazon.science/blog/bridging-intent-and-execution-in-agentic-systems","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/real-world-grounding-in-agentic-ai","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/building-trust-into-ai","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/building-trust-into-ai","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/formally-verified-aes-xts-the-first-aes-algorithm-to-join-s2n-bignum","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/news/how-academic-collaboration-delivers-real-world-security-to-amazon-customers","path":null,"label":"amazon.science/news","type":"external"},{"url":"https://www.amazon.science/blog/making-llms-faster-without-sacrificing-accuracy","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/making-llms-faster-without-sacrificing-accuracy","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/intelligence-isnt-about-parameter-count-its-about-time","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/intelligence-isnt-about-parameter-count-its-about-time","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/optimizing-lora-target-module-selection-for-efficient-fine-tuning","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/optimizing-lora-target-module-selection-for-efficient-fine-tuning","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/promptimus-improving-already-good-llm-prompts-with-zero-manual-engineering","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/promptimus-improving-already-good-llm-prompts-with-zero-manual-engineering","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/diverse-reasoning-traces-teach-llms-to-make-better-decisions","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/diverse-reasoning-traces-teach-llms-to-make-better-decisions","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/how-mechanism-design-theory-helps-optimize-amazon-vendor-collaboration","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/why-a-12-year-old-forecasting-paper-has-stood-the-test-of-time","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/news/aws-gray-lab-johns-hopkins-announce-groundbreaking-database-for-ai-ml-antibody-design","path":null,"label":"amazon.science/news","type":"external"},{"url":"https://www.amazon.science/news/aws-gray-lab-johns-hopkins-announce-groundbreaking-database-for-ai-ml-antibody-design","path":null,"label":"amazon.science/news","type":"external"},{"url":"https://www.amazon.science/blog/improving-quality-and-robustness-in-llm-based-text-to-speech-systems","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/improving-quality-and-robustness-in-llm-based-text-to-speech-systems","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/how-ai-is-changing-the-nature-of-mathematical-research","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://www.amazon.science/blog/ground-truth-is-a-process-not-a-dataset","path":null,"label":"amazon.science/blog","type":"external"},{"url":"https://aws.amazon.com/blogs/machine-learning/evaluate-your-amazon-nova-sonic-voice-agent-at-scale-no-microphone-required/","path":null,"label":"aws.amazon.com/blogs","type":"external"},{"url":"https://www.amazon.science/research-awards/latest-news/fall-2025-amazon-research-awards-recipients-announced","path":null,"label":"amazon.science/research-awards","type":"external"},{"url":"https://www.amazon.science/research-awards/latest-news/fall-2025-amazon-research-awards-recipients-announced","path":null,"label":"amazon.science/research-awards","type":"external"}],"provenance":{"provider":"deepseek","model":"deepseek-v4-pro","workflow":"onlylabs-deepagents-analysis-v3","agent":"deepagents"},"evidence":{"total":92,"pages":28,"events":140,"web":4,"signal_desks":{"forks":0,"repos":19,"hiring":0,"talking":22,"releases":19},"data_radar_lanes":{"data":10,"evals":6,"safety":5,"product":2,"infrastructure":8},"data_radar_matches":21}},"signal_counts":{"total":228,"model_released":30,"release":33,"repo_new":136,"repo_forked":0,"post_published":29,"job_opened":0}}