prof_pic.jpeg

Tai Dang

I am an AI Resident at Xaira working on Generative Diffusion Models and Reinforcement Learning, with applications in AI for Science, particularly computational biology and molecular discovery.

My research focuses on developing principled generative models and reinforcement learning methods for scientific discovery at scale.

Selected Research

* denotes equal contribution

  1. NeurIPS’26
    Plausible Biomolecular Structure Prediction via Physics-informed Reinforcement Learning
    Tai Dang, Hieu Tran, Long-Hung Pham, and 4 more authors
    Paper Blog Code
  2. ICML WS’25
    Preferential Multi-Objective Bayesian Optimization for Drug Discovery
    Tai Dang*, Long-Hung Pham*, Sang T. Truong*, and 6 more authors
    Paper Code
  3. ICML WS’26
    Gathering Context that Supports Decisions via Entropy Search with Language Models
    Sang T Truong, Sicong Huang, Pranava Singhal, and 6 more authors
  4. EACL’23
    Enriching Biomedical Knowledge for Low-resource Language Through Large-Scale Translation
    Long Phan*, Tai Dang*, Hieu Tran*, and 4 more authors
    Paper Code

Experience

Xaira Therapeutics — AI Resident
2026

  • De Novo Design Team

Stanford University — Visiting Researcher
2024 – 2026

  • Post-trained AlphaFold 3 via reinforcement learning, achieving SOTA structure fidelity.
  • Optimized large-scale drug screening with Bayesian optimization.

University of Massachusetts Amherst — Research Assistant
2023

  • Built multi-modal retrieval system for Outside-Knowledge Visual QA.
  • Analyzed policy specialization in OLS Convex Coverage Set under varying discount factors.

Ontocord — Research Intern
2023

  • Distilled a 7B LLM to 5× smaller size while maintaining performance.
  • Built open-source Vietnamese LLM using 1TB processed data.

EOG Resources — Software Engineer Intern
2023

  • Built graph-based visualization tools for complex data analysis.
  • Migrated repositories to GitHub Actions with OIDC authentication.

VietAI — Research Intern
2022

  • Developed SOTA English–Vietnamese translation model.
  • Improved biomedical NMT by +6 BLEU and released Vi-MedNLI dataset.

Education

Stanford University
Visiting Researcher
Advisor: Thang Luong (Google DeepMind), Jeff Glenn (Stanford Medicine)

University of Massachusetts Amherst
B.S. in Computer Science, May 2024
Advisor: Bruno Castro da Silva