Anton Morgunov

Yale University

Advisor: Prof. Victor S. Batista

Yale University

Grade Average: Honors

Massachusetts Institute of Technology

GPA: 5.0/5.0. Phi Beta Kappa.

Prof. Victor S. Batista Lab, Yale University

• Developed DirectMultiStep, a Transformer-based retrosynthesis model (Mixture-of-Experts) that eliminates search-space explosion, achieving 3.1x higher accuracy than graph-search methods.
• Secured $100k prize (2nd Place) in the Standard Industries Chemical Innovation Challenge by deploying these generative models to solve industrial synthesis targets.
• Engineered ChemSpaceAL, an active learning framework using latent-space upsampling that reduced the computational cost of alignment by 90% (10x efficiency gain).

Prof. Troy Van Voorhis Lab, MIT

• Developed a composite electronic structure method recovering Coupled-Cluster (CCSD) accuracy within 0.02 eV at a fraction of the computational cost (MP2 scaling).
• Automated large-scale benchmarking pipelines (2k+ lines Python/Shell), streamlining the calculation → analysis → visualization workflow for high-throughput spectroscopy prediction.

Prof. Ronald T. Raines Lab, MIT

• Executed multi-step organic synthesis of novel diazo compounds for protein bioconjugation, characterizing products via NMR and MS.
• Bridged experimental results with theory by modeling transition states using DFT to elucidate reaction mechanisms.

Stealth Startup

• Architected end-to-end reaction plausibility engine and negative data generation pipeline for proprietary workflows.
• Refactored GNN codebase into production-ready PyTorch, resolving bottlenecks to achieve 2x inference speedup.
• Engineered automated benchmarking suite to validate model performance against internal experimental datasets.

Los Alamos National Laboratory

• Modernized legacy infrastructure by porting MATLAB codebases to high-performance Python/VASP pipelines.
• Identified and fixed critical theoretical discrepancies in published methodology, restoring reproducibility to a stalled research project.

Universal Benchmarking Infrastructure

• Built the industry-standard evaluation harness (RetroCast) and visualization engine (SynthArena), solving the field's data heterogeneity problem to enable the first rigorous comparison of generative vs. search-based planners.
• Sole architect of full-stack system (Next.js, Python, SQLite); engineered cryptographic provenance tracking and deployed via self-managed CI/CD pipelines.

models.batistalab.com

• Engineered and maintain a full-stack MLOps platform serving models from 5 publications; currently processing live prediction traffic for 130+ external researchers.
• Managed bare-metal GPU infrastructure (VPS/Docker/NGINX) to achieve zero vendor lock-in and data sovereignty.

Beyond Curriculum Public Foundation

• Built and managed a remote organization of 70+ volunteers; scaled platform to 2M+ monthly pageviews serving 190k+ unique students.
• Secured over $40,000 in grants and corporate sponsorship to democratize STEM education access in rural regions.

Kazakhstan Chemistry Olympiads Association (QazChO)

• Implemented data-driven selection pipelines that resulted in 5 consecutive years of IChO gold medals (ending a 5-year drought) and represented national delegation as arbitrator at IChO 2019-2022.