Senior Research Engineer at American Express, working on agentic AI systems.

I'm exploring the intersection of search algorithms and language models, particularly using Monte Carlo Tree Search (MCTS) to improve AI reasoning capabilities. My work centers on making smaller models more capable through intelligent orchestration rather than simply scaling parameters.

• MCTS for RL Training Data: Using tree search methods to generate high-quality training data for reinforcement learning
• Model Distillation: Developing techniques to compress large model capabilities into smaller, deployment-friendly versions
• MCP Servers: Building Model Context Protocol servers to bridge research implementations with production systems
• Conversational Optimization: Applying search algorithms to explore conversation spaces and find optimal response strategies

Languages: Python, TypeScript, C++
ML Frameworks: PyTorch, vLLM, TensorFlow
Infrastructure: LangGraph, MCP, FastAPI

I believe the path to more capable AI systems lies not just in larger models, but in better search algorithms and tool orchestration. A well-designed 7B parameter model with proper tooling can often outperform a naive 70B model on real-world tasks.

Started in NLP despite being advised against it by a professor in 2021—turned out to be the best advice I never took. Since then, I've built multi-agent frameworks serving thousands of users in production environments.

Conversational Analysis Engine - An MCTS-based system for optimizing human-AI conversations by exploring multiple response paths before selection.

Feel free to reach out if you're interested in search algorithms, agentic systems, or making AI more practical for real-world applications.