I specialize in building algorithms for high-precision optical metrology, and I love coding models from scratch to understand them under the hood. I build systems across NLP, RL, and Generative AI.
- Exploring reinforcement learning, agentic AI, world models, and memory systems
- Writing about my learnings and projects at TensorWrites
- Read more about my experience in my resume
- Check my reading list here
- Movies, Series, and Sitcoms I Have Watched
| Project | Description |
|---|---|
| AresSim | A Mars habitat development environment for testing reinforcement learning algorithms and LLM agents. Agents control a rover and perform tasks involved in building and maintaining a habitat. |
| LLM Gathering | A platform where multiple LLMs with assigned roles discuss a given topic. It explores multi-agent reasoning and memory systems for collaborative agent workflows. |
| Project | Description |
|---|---|
| Trisynapse | Trisynapse is a local-first, BYOK AI research workspace. Turn scattered papers, PDFs, bookmarks, notes & videos into a private digital brain. The app auto-summarizes, maps concept graphs, and compiles sources into auto-updating wikis. Work seamlessly with a built-in split-view Browser, local retrieval Chat, AI Assistants, and infinite Canvas whiteboards for note-taking. 100% offline-first (except AI API calls) and private: your vault stays on your machine |
| DotChat | A minimalist PyTorch framework for training small language models from scratch. It supports pretraining, SFT, DPO, Chain-of-Thought reasoning, CUDA Graphs, and static KV-cache inference. |
| SimpleThoughts | A synthetic Hugging Face dataset designed for supervised fine-tuning, preference alignment, and reasoning training. |
| ArxiViz | An interactive platform for exploring neural network architectures across computer vision, NLP, generative AI, and reinforcement learning alongside mathematical concepts and PyTorch implementations. |
| ScratchNLP | A Python library implementing core NLP algorithms and tokenizers entirely from scratch. |