📖 About Me
🧹 Building Sweep, an AI powered junior dev🔗 Find my info at https://kevin-lu.tech👤 Pronouns: He/Him
🔭 I previously worked on ...
📝 Aline: a tool to reshape research with context-aware auto-generation📕 PyATE: Python Automated Term Extraction, a Python package that implements five automated term extraction (ATE) algorithms based on eight research papers, attaining over 20,000 downloads!♻️ Recycler: Fine-tuned model of Google's XCeption that uses image recognition to classify images of recyclables by materials, such as glass or metal, reaching 86% accuracy, using Gary Thung's dataset- ✏️ Research Mode: Chrome Extension that provides a convenient sidebar for note-taking as well as other NLP-based utilities for accelerating research, such as automated jargon highlighting, text simplification and reader mode. Future features include auto-generated citations, semantic search and automated summarization.
🌱 I’m currently learning ...
- Locally running LLMs for code generation (CodeGen-2.5-mono + ONNX maybe?)
- Training and evaluating better natural language code search embedding models
- Interested in fine-tuning models to act as agents who can use tools. Maybe reinforcement learning on successful/unsuccessful HotPotQA runs?
- Using language models to do static asset manipulation, such as generating my favicons for my site
- Gallian's Contemporary Abstract Algebra
- Stein and Shakarchi's Princeton Lectures in Analysis II: Complex Analysis
🤔 I’m looking for help with ...
- Evaluating and fine-tuning GTE on natural language to code search, such as CodeSearchNet. It's amazing at most things but not evaluated on code search.
🎊 Fun fact
📫 How to reach me ...
📧 Email: kevinlu1248@gmail.com🐦 Twitter: @kevinlu1248
Credits
- Raymond Li for his intro gif; I forked it for my own needs
- Brittany Chiang for her open-sourced personal website; I used it as a Gatsby template to create my own personal website