Lucy Zimmerman
Full-Stack AI Builder & Researcher
I'm an AI engineer and Stanford CS master’s student specializing in applied machine learning, AI safety, and product engineering. My background blends full-stack development and AI research — I'm currently building Sub-Line.com, an AI system that turns decades of news archives into searchable, interpretable intelligence for journalists and researchers — supported by a $125K Brown Institute for Media Innovation Magic Grant. I’m excited by roles that connect cutting-edge AI tools and research to products with real impact, especially in human-AI collaboration.
About
I'm passionate about building AI systems that are reliable, interpretable, and human-centered. My work spans both AI research and full-stack engineering, from fine-tuning custom NLP models to designing interactive platforms for human oversight.
Currently, I'm working on SubLine, a full-stack AI platform for journalists featuring custom NLP fine-tuning, an interactive graph UI, and human-in-the-loop annotation systems.
My past work includes TransparencyGPT (bias detection app selected for Dorm Room Fund), research at Stanford HAI on AI harms, and computer vision + geospatial ML at the Human Trafficking Lab.
Core philosophy: AI must be reliable, interpretable, and human-centered to create meaningful impact for society.
What I'm Working On
SubLine – AI Research Platform for News Organizations
SubLine is a full-stack AI application designed to help journalists explore complex information through AI-powered knowledge graphs.
- Custom NER Model: Trained domain-specific transformer-based entity extraction model to capture noisy and local entities
- Document-Level Relation Extraction: Built comprehensive LLM fine-tuning pipeline on RE-DocRED with multi-label classification across 97 relation types (Mistral-7B, RoBERTa, DeBERTa)
- End-to-End Deployment Platform: Engineered complete system with preprocessing, automated evaluation metrics, HuggingFace Hub integration, and production-ready deployment
- Human-in-the-Loop Graph Interface: Built interactive React-based knowledge graph editor where reporters highlight text, review entities/relationships, and approve modifications that flow back into the training loop
- Active Pilots: Currently running pilots with newsrooms to refine the platform
Tech Stack: PyTorch, HuggingFace Transformers, React, Next.js, Flask, Supabase
Projects
Skills
Full-Stack AI Applications
Deployment & Engineering
Core ML & Research
Beyond Code
Timeline
AsylumAssist
Founding Full-Stack AI Engineer building mobile platform to guide asylum seekers through legal process.
SubLine
Leading development of full-stack AI platform for newsrooms.
TransparencyGPT
Built explainable bias detection app, selected for Dorm Room Fund.
Stanford HAI
Contributing to AI Index harms chapter, bridging technical work and policy.
Human Trafficking Lab
Built ML pipeline for satellite imagery + trafficking detection.
minBERT/SimCSE Research
Improved embeddings for multiple NLP downstream tasks.
Experience
Founding Full-Stack AI Engineer
Developing a mobile-first platform with the legal team at the Interfaith Center in NYC that guides asylum seekers through the court process and helps them generate legally-sound declaration PDFs with LLMs. Impact: Reduces prep time from weeks to hours, saving $5k–$15k per case.
Co-Founder & Lead Engineer
Building an AI platform for news organizations. Leading custom NLP model development, human-in-the-loop annotation platform, and interactive knowledge graph UI.
CS + Social Good Fellow
Lead Research Intern for Senior Advisor for Technology and Innovation focused on AI + Human Rights.
platform
Stanford Human Trafficking Data Lab
Implemented remote detection algorithms leveraging satellite and geospatial data for pinpointing trafficking hotspots. Collaborated on computer vision models, contributing to ethical applications of AI in global challenges.
User platform, Product and Engineering Team
Conducted 50+ user interviews to assess product fit and prioritize engineering solutions for harassment prevention tools. Organized beta testing initiatives and recruitment campaigns, driving iterative improvements in user experience.
platform
Stanford Human-Centered Artificial Intelligence Center
Delivered 50+ data-driven analyses featured in the 2023 AI Index report, highlighting AI's societal impacts. Explored constraints within AI systems to propose strategies for fostering inclusivity in AI applications.
Software Engineer
Modernized software architecture for logistics platforms, coding over 2,000 lines and integrating scalable infrastructure. Supported operational systems for top global freight forwarders, improving efficiency and reliability.
Product Engineer
Built Blindsight, a safety solution for industrial environments powered by AI vision.