
Harry Li
I am an XAI and VIS researcher and full stack web developer at MIT Lincoln Laboratory. I love to code solutions for problems I see around me!
MENU
Appetizers
Education
Tufts University
May 2023
Master's in Computer Science
Courses included: Network Security, Offensive Security, Deep Learning, Graph Neural Networks, Reinforcement Learning
Master's Project, presented paper at IEEE VIS 2023: Knowledge Graphs in Practice: Characterizing their Users, Challenges, and Visualization Opportunities
Rutgers University
May 2018
Bachelor of Science, Electrical and Computer Engineering
Economics Minor
Engineering Honors Academy, High Honors
Entrees
Work and Research
MIT Lincoln Laboratory
2018 – Present
Associate Staff, 2021 – Present
Assistant Staff, 2018 – 2021
AI Technology and Systems Group
- Principal Investigator / Project Lead
- Developing LLM-assisted user inference for knowledge graph (KG) data exploration and question answering
- Created prompting pipeline for LLM to directly explore KG data
- Mitigates hallucinations with ground-truth KG data
- IEEE VIS 2024 Paper: LinkQ: An LLM-Assisted Visual Interface for Knowledge Graph Question-Answering
- Principal Investigator / Project Lead
- Researching uncertainty quantification in machine learning
- Developed scatter plot visualization to communicate model uncertainty for multiclass classification and to identify out-of-distribution test samples
- IEEE VIS 2023 Paper: ScatterUQ: Interactive Uncertainty Visualizations for Multiclass Deep Learning Problems
- VizSec21 Best Poster: Interactive Visual Uncertainty Quantification
Desserts
Projects
LobbyView
Created an interactive data visualization website to explore US government lobbying data
Frontend Technical Lead
StyleGAN2 and Latent Space Discovery Techniques
Created novel diamond and gemstones datasets
Trained StyleGAN2 to generate realistic diamonds and gemstones
Used latent space discovery techniques to control output features
In-Browser Object Detection with TensorFlow.js
Created a website that uses SSD Mobilenet to do object detection completely in the browser with TensorFlow.js
Svelte Kit and TypeScript
End-to-End Encrypted React - Rust Chat
Created an end-to-end encrypted messaging system using React and Rust
Reinforcement Learning Pac-Man
Trained an RL agent to play Pac-Man
Tree-search state-action encoding
N-Step Semi-Gradient SARSA
Linear function approximation and gradient descent
Drinks
Skills
Front End
TypeScript
React
React Native
Redux
Svelte
D3.js
Bootstrap
Mantine
Back End
Express.js
Flask (Python)
Next.js
Database
MongoDB
PostgreSQL
SQLite
Testing
React Testing Library
Vitest
LLMs
OpenAI / GPT
LLAMA
Mistral
Misc
Auth0
AWS
Figma
Git
NumPy
PyTorch
Serverless
Stripe
Tensorflow
Reservations
Email: harry.li at ll.mit.edu