Yuntao Zhang

Welcome! I am a Senior Scientist at Bristol Myers Squibb based in the San Francisco Bay Area, specializing in high-throughput mass spectrometry, molecular modeling, and AI-driven drug discovery. My work bridges experimental bioanalytical chemistry and modern computational toolchains—integrating high-resolution structural mass spectrometry with foundation models to accelerate biologics discovery, automate analytical pipelines, and elucidate biomolecular mechanisms.

About Me

• Current Role: Senior Scientist at Bristol Myers Squibb (BMS)
• Expertise:
  • Biologics Mass Spectrometry: Characterizing biologics with peptide mapping, carbene footprinting (epitope/paratope mapping), and bottom-up proteomics.
  • Computational & AI: Applying molecular modeling, deep learning, and AI agents to solve complex scientific problems.
  • Workflow Automation: Developing automated systems to enhance research efficiency and reproducibility.
• Education:
  • Ph.D. in Bioanalytical and Physical Chemistry, University of the Pacific
  • M.S. in Environmental Engineering, University of the Pacific
  • B.S. in Ecology, University of Science and Technology Beijing

Featured Interactive Project

MoE 3D Architecture Visualizer

Three.js / WebGL React 18 Sparse MoE 535B LatentMoE Compression Dynamic Sampling

An open-source interactive 3D WebGL visualization platform for modern 500B+ sparse Mixture-of-Experts (MoE) LLM architectures, featuring LatentMoE communication compression, Quantile-Balanced routing, and real-time dynamic sampling pipelines running entirely in the browser without server dependencies.

🚀 Live Demo 💻 GitHub 📝 Technical Blog 🔍 Details

Unique Skillset

With a strong foundation in analytical chemistry and a deep interest in computer science and AI, I have developed a unique interdisciplinary profile. My work combines hands-on expertise in mass spectrometry and biologics with advanced computational approaches. I specialize in fine-tuning and applying state-of-the-art foundation models—such as ST-Tahoe for single-cell perturbation prediction and Boltz-2 for structural biology—to solve complex scientific challenges. My skills span the full MLOps lifecycle, from setting up and managing GPU-powered training environments on GCP and Domino Data Lab to employing modern development tools like Docker and UV for reproducibility. This enables me to bridge the gap between cutting-edge AI and practical scientific discovery, from high-throughput data processing to generating deep biological insights.

Complementing my scientific modeling expertise, I maintain active technical fluency in foundational computer science and algorithms. I regularly explore algorithmic implementations across multiple programming languages—including C++, Go, Java, JavaScript, and Python—to evaluate diverse concurrency models, memory architectures, and runtime execution paradigms, strengthening my ability to engineer high-throughput scientific pipelines.

Agentic Workflow

Leveraging autonomous CLI agent workflows (Claude Code, Gemini CLI, GitHub Copilot) for rapid full-stack prototyping, scientific pipeline automation, and reproducible research.

Impact & Achievements

• Open-Source AI Architecture Visualization: Engineered client-side WebGL rendering and communication compression algorithms for 500B+ sparse Mixture-of-Experts (MoE) architectures in MoE 3D Architecture Visualizer, enabling real-time interactive routing and dynamic sampling directly in the browser (Demo / Code).
• Advanced Model Fine-Tuning: Systematically fine-tuned the ST-Tahoe foundation model for gene perturbation analysis, diagnosing and resolving complex issues like representation mismatch and optimizer state conflicts to significantly improve performance. This work involves deep analysis of model architecture, loss functions, and training dynamics on enterprise-level platforms.
• Biologics Characterization: Led the implementation of carbene footprinting for high-resolution epitope/paratope mapping, and developed novel LC-MS methods to support bispecific, ADC, and CAR-T programs.
• Mentorship & Collaboration: Trained and mentored scientists in advanced MS and computational techniques. Recognized with the CABS 2026 Mentor Impact Award for exceptional guidance and dedication in the CABS Data Science Summer Internship Program.

Publications & Research

I have authored multiple peer-reviewed publications in leading journals, advancing the field of mass spectrometry, peptide/peptoid analysis, and molecular modeling. My research is rooted in fundamental physical chemistry, exploring the interplay between molecular structure, chirality, and analytical measurement. Bridging gas-phase ion chemistry, high-resolution mass spectrometry, and quantum chemical modeling to elucidate biomolecular structure and reactivity. See my publications for details.

Professional Development & Certificates

I am committed to continuous learning in scientific programming, AI, and leadership. I enjoy building innovative tools and exploring new technologies. I hold certificates in:

• Machine Learning (Stanford/Coursera)
• Deep Learning Specialization (DeepLearning.AI)
• AI Agents (Hugging Face)
• Associate Python Developer (DataCamp)

Explore all certificates and projects.

Vision

I believe the next frontier in biotherapeutics lies at the intersection of high-resolution bioanalytical mass spectrometry and foundation AI models. By uniting rigorous experimental measurements (such as covalent carbene footprinting) with generative and self-supervised architectures, my vision is to accelerate the rational design of targeted therapeutics, automate complex analytical discovery, and uncover molecular mechanisms underlying human aging and complex diseases.

Feel free to explore my CV, publications, projects, and certificates to learn more about my work. I am always open to new collaborations and opportunities to make a positive impact in the scientific community.