π€ Senior Research Engineer @ American Express | Building the future of agentic AI, one search tree at a time
- Search + LLMs = Magic β¨ Currently using MCTS to teach AI emotional intelligence through self-play
- Small models, big brains π Why use a 70B model when a 7B model with clever orchestration and a whole host of tools might work better?
- Agentic systems that actually ship π Built multi-agent frameworks serving thousands
- π― Monte Carlo Tree Search for generating RL training data
- π Distilling large models into lean, mean, reasoning machines
- π οΈ MCP servers that turn academic papers into production code
- π³ Teaching AI to explore conversation trees and find optimal paths
Languages: Python, TypeScript, C++ | ML: PyTorch, vLLM, TensorFlow | Orchestration: LangGraph, MCP, FastAPI
- Told by a professor to avoid NLP in 2021 (best advice I never took π)
- Firm believer that search algorithms + tool use = AGI-lite
Conversational Analysis Engine - MCTS meets LLMs to optimize human conversations. Because sometimes the best response isn't the first one you think of.
"The future of AI isn't just bigger modelsβit's smarter search and better orchestration" π―
π« Let's connect and build something awesome together!