Thy Tran
AI Engineer – Scientific Software Engineer | Life Sciences & ML | MS in Software Engineering Systems
Hire this person
Send a job offer directly to this candidate
About
ML engineer with a real-time graphics and C++ systems background. Created a spatial protein visualization platform processing 216M+ variants (30x query speedup) and a multi-agent AI system at Moderna that cut configuration time by 95%. Applying point cloud deep learning, neural reconstruction (NeRF, Gaussian Splatting), and computer vision to bridge rendering and scene understanding.
Experience
SDE Co-op, Systems Integration, Applied Tech at MODERNA, INC (2025-07 – 2025-12)
- Developed an intelligent multi-agent system that automated laboratory configuration generation for Moderna's automation workflows, generating configuration files for Hamilton Liquid Handler batch jobs, reducing configuration time by 95% (from 30+ minutes to 2 minutes) while achieving 90-95% generation accuracy across four configuration types.
- Architected a stateful orchestration system using LangGraph with GPT-4o that routes conversations through specialized builder agents, implementing a cost-optimized 3-stage hybrid intent detection pipeline (heuristics → JSON parsing → LLM fallback) and conversational JSON editing with real-time Pydantic validation.
- Built FastAPI backend with async architecture and SQLite-based session persistence for multi-session state management, enabling modular parallel processing through specialized builder agents.
- Designed schema-guided configuration generation with file artifact versioning, allowing non-technical users to generate expert-level configurations through natural language prompts.
- Delivered high consistency in configuration structure and naming conventions, significantly reducing human error rates while democratizing access to configuration generation and eliminating the need for specialized domain expertise.
Graduate Teaching Assistant at NORTHEASTERN UNIVERSITY (2024-09 – Present)
- Mentor 15 graduate students in C++ OOP including advanced topics (Templates, Smart Pointers, Move Semantics)
- Guide students in data structures and algorithms implementation, achieving 20% average exam score improvement
Software Engineer at BOSCH GLOBAL SOFTWARE TECHNOLOGIES (2021-08 – 2023-05)
Diagnostic Fault-Tolerant Tool (DFC)
- Built Python/PyQt5 diagnostic data automation tool and TypeScript/Angular FMEA risk assessment module with API integration, achieving 30-40% efficiency improvements in automotive quality workflows
Education
Master of Science in Software Engineering Systems (Thesis Track) – NORTHEASTERN UNIVERSITY (2024-01 – 2026-05)
