I actively contribute to open source projects in AI agents and AI-assisted tools for chip design. My work spans from production AI agent runtimes to MCP servers for developer tooling.
Open Source Projects
nova-agent-runtime — Python AI agent framework with integrated tools, multi-agent workflows, and evaluation loops. Features native MCP integration, automatic context management, and YAML-based declarative configuration.
openroad-mcp — Python, Tcl MCP server for real-time AI-driven debugging for chip designers in OpenROAD.
ORAssistant — Python, Tcl Production-ready LLM chatbot assistant for the OpenROAD chip design tool. Conceptualised and open-sourced as part of OpenROAD Project contributions.
Publications
ORAssistant: A Custom RAG-based Conversational Assistant for OpenROAD (WOSET 2024) Paper on a custom RAG-based conversational assistant for chip design automation using OpenROAD.
Extended Tactile Perception: Vibration Sensing through Tools and Grasped Objects (IROS 2021 — Best Paper Award) ML classification models for high-dimensional time-series tactile sensor data enabling vibration sensing through tools and grasped objects.
Talks
Free Silicon Conference 2026 (upcoming) Talk on openroad-mcp — title TBC.
Writing
OpenROAD MCP: Let AI Close Your Timing (Feb 2026) Introducing OpenROAD MCP — a server that enables AI assistants like Claude Code to interact directly with OpenROAD processes for timing analysis and constraint adjustment.
Building Robust AI Research Agents: A tale of five specialists (Nov 2025) Four architectural iterations of a deep research workflow for VCs — from a single agent to five specialised agents with hierarchical citations, MCP integration, and iterative refinement.
Multi-Agent Systems Are 23x Cheaper Than Deep Research Models with Higher Quality (Nov 2025) Benchmarked multi-agent systems (11 specialised agents) against single-model and deep reasoning models across 5 scientific domains. Multi-agent won or tied in all 5 at 23× lower cost.