Frontier AI Paper Reading Group: Post-NeurIPS 2025

​Post-NeurIPS 2025: Paper Reading

​🔬 Topics

​⚡ Scaling laws & emergent capabilities

​🛠️ Production-ready techniques

​💡 Research → startup opportunities

​🔍 What's actually new vs hype

​🎙️ Speakers

​🤖 Jonathan Li Caltech • Asari AI • NEC Labs Scaling LLM reasoning through reinforcement learning and inference-time search—applications in code generation, computer-use agents, and multi-agent systems.

​🛡️ Junyu Guo PhD, IEOR • UC Berkeley DRCORL's approach to the utility-safety tradeoff in offline learning, enabling agents to follow constraints without sacrificing reasoning capability.

​🎭 Yi-Hsuan Tsai Co-Founder/CTO • Atmanity Avatar personalization breakthrough preserving identity-specific facial details using lightweight model adapters.

​🧬 Mia A. Rosenfeld, PhD Iambic Therapeutics • Protein-ligand structure prediction in drug discovery—computational approaches for molecular binding predictions in therapeutic development.

​⚙️ Chinmay Deshpande Amazon Lab126 • Interactive AI Infrastructure Design Tool for architecture decisions—serving strategies, KV cache, framework selection, and parallelism—with benchmarks, cost calculations, and exportable diagrams from 20+ research papers.

​🤝 Co-Hosts

​🌐 The AI Collective aicollective.com

​🧪 Chemistry chemistry.vc

​🏔️ AI Valley aivalley.io

​💼 Diversis diversis.com