Gregory W. Kyro, PhD

About

I am a Scientist on the AI Research Team at Lila Sciences, where I develop and lead technical initiatives at the intersection of LLM post-training and experimental science—enabling Lila’s central AI agent to learn from proprietary laboratory data, experimental workflows, and feedback from the physical world.

Selected work
  • Architected Lila’s >10-trillion-token data generation engine that produced the foundation of Lila’s post-training corpus. Created a scalable system for converting experimental data into natural-language scientific tasks with programmatic verification, generating more than 10 trillion high-quality scientific tokens that now form the foundation of Lila’s post-training corpus. The system is the subject of a patent application on which I am the lead inventor.
  • Orchestrated Lila’s first lab-in-the-loop AI recursive self-improvement cycle. Spearheaded the cross-functional effort that enabled Lila’s first lab-in-the-loop scientific self-improvement cycle, in which the company’s central AI agent proposed experiments, received measurements from an automated laboratory workflow, and used the resulting feedback to iteratively refine subsequent experimental designs—establishing Lila’s first closed-loop implementation of scientific self-play against physical reality.
  • Built Lila’s company-wide framework for translating domain-specific data, workflows, and intelligence into reinforcement learning environments. Developed Lila’s shared framework for transforming domain-specific proprietary data, real laboratory workflows, and scientific objectives into reinforcement learning environments used to train the company’s central model. I now lead its implementation throughout cross-organizational scientific AI projects to ultimately drive distributed domain-specific efforts toward coherent programs that advance the scientific capabilities of Lila’s AI agent.
  • Led post-training research for scientific reasoning and generalization. Drove multiple research directions in LLM post-training to improve scientific reasoning and cross-domain generalization, contributing to the design of Lila’s post-training program.

Prior to Lila, I earned my PhD in Computational Biophysical Chemistry from Yale University in three years as an NSF Fellow, developing AI models that advanced the state of the art in drug discovery and were implemented in collaboration with leading companies including Pfizer, NVIDIA, Novartis, Moderna, Merck Group, and others.

Selected research
  • HAC-Net: AI model for predicting protein-ligand binding affinity—used to identify therapeutic candidates for cancer, diabetes, multiple sclerosis, and drug-resistant infections; and to enable a prostate cancer imaging platform.
  • CardioGenAI: generative AI framework for re-engineering drugs for reduced hERG-related cardiotoxicity—implemented at Pfizer R&D; evolved into SafeSynthAI—first-place winner of the Blavatnik Fund for Innovation at Yale Ventures, seeding a spinout now in development.
  • T-ALPHA: successor to HAC-Net; AI model for predicting protein-ligand binding affinity—used to identify therapeutic candidates for thoracic aortic aneurysm.
  • ChemSpaceAL: active learning method for target-conditioned molecular generation—used to design small-molecule binders to CRISPR-Cas9 to enhance its specificity for target DNA sequences.
  • Electrostatic Eigenvector Centrality: spectral graph-theoretic method for describing intra-protein information transfer—used to map experimentally validated allosteric mechanisms in CRISPR-Cas9, IGPS, and MIF-2.

Concurrently with my doctoral research, I drove scientific initiatives that drew millions of dollars in government funding, developed technology adopted across the pharmaceutical industry, founded and led scientific organizations, advanced frontier scientific-AI efforts across leading technology and life-science companies, and earned broad external recognition for scientific achievement and leadership—winning prestigious competitions, receiving distinguished awards, and being featured across major scientific media outlets.

Selected impact
  • Led the foundational science that catalyzed a $10M state investment in a quantum initiative of which I was a founding scientist. Served as one of the founding scientists of QuantumCT—a Yale-affiliated initiative aimed at accelerating quantum research partnerships between public and private sectors—where I led the foundational scientific efforts, established collaborations with numerous industry partners including Pfizer, NVIDIA, Novartis, and Moderna, and published the initiative’s first collaborative paper as first author—efforts that ultimately catalyzed a $10M investment from the Connecticut State government.
  • Co-developed SAGE—an AI-based rapid-synthesis framework that was acquired by Merck Group and is currently being integrated across multiple teams within the company.
  • Founded and served as President of the Yale University Chapter of the Biophysical Society, and mentored researchers who subsequently joined xAI, NVIDIA, and the lab of Nobel laureate David Baker.
  • Won first place at highly selective competitions including the Merck Innovation Cup, with awards from leading federal science agencies such as the National Science Foundation and the National Institutes of Health—garnering coverage across major scientific media outlets including Yale News, the Yale Alumni Magazine, and the Biophysical Society.
  • Published 18 papers in top-tier journals including Chemical Reviews, Nature Computational Science, and Nature Communications; delivered more than 30 talks, lectures, panels, and conference presentations; and released multiple open-source scientific software packages.
  • Advanced frontier scientific-AI efforts across OpenAI, NVIDIA, and Moderna—contributing expert scientific data annotation to an OpenAI LLM post-training effort through Scale AI, designing quantum machine learning methods in collaboration with NVIDIA and Moderna, and developing a PROTAC virtual screening pipeline at OpenEye, Cadence Molecular Sciences.
  • Built a network of >10 research collaborations across leading pharma companies and top academic labs—spanning Drug Safety & Toxicology at Pfizer; Quantum Algorithm Engineering at NVIDIA; Computational Protein Design & Modeling at Moderna; Preclinical Safety at Novartis; Computational Antibody & Protein Engineering at Boehringer Ingelheim; the Schwartz (Cardiology), Hafler (Neuroimmunology), and Kaminski (Pulmonology) Labs at the Yale School of Medicine; the Lisi Lab (Molecular Biology & Biochemistry) at Brown University; the Palermo Lab (Bioengineering) at UC Riverside; and the Loria Lab (Chemistry & Biophysics) at Yale University.