Hi, I'm
Building at the intersection of machine learning and silicon.
EECS Alum at UC Berkeley with a passion for the intersection of AI and hardware. My work spans agentic systems for hardware design & verification and LLM agent research.
Things I've worked on over the years
Led end-to-end development of an LLM-powered tool using graph theory to analyze and auto-resolve clock power gating failures in AF. Reduced debug turnaround from 5 days → 5 minutes.
A novel architecture for LLM agents that achieves low-latency, interactive performance by decoupling I/O from inference and speculatively executing tool calls before the model commits to them. Research conducted at Berkeley AI Research (BAIR).
Built verification infrastructure and verification agents from scratch for an ASIC (Atlas NPU) designed to accelerate VLA physical intelligence models, taped out on TSMC's 16nm process. Team was awarded Best Student Poster at the 2026 UC Berkeley SLICE Lab retreat.
Built a RAG-based copilot with LangChain to explain, debug, and generate TCL scripts for analog circuit simulations. Fine-tuned ColBERT to 98% retrieval accuracy. Deployed company-wide to 9,000+ engineers — replacing a legacy automation team.
Apple SoC
Research on interactive real-time LLM agents using asynchronous I/O and speculative tool calling — published on arXiv. Advised by Josh Kang.
Led development of SimSight using LLMs and graph theory to resolve clock power gating issues in Apple Fabric SoCs. Built advanced LLM chains for failure categorization and signal analysis on waveforms.
Designed a RAG-based LLM copilot for TCL script generation and debugging. Fine-tuned ColBERT for 98% retrieval accuracy. Integrated into Synopsys' internal AI assistant, used by 9,000+ analog engineers globally.
Cultivated industry and academic partnerships with Synopsys, Kibsi, Bright Machines, and Lightning AI. Managed projects worth 6+ figures cumulatively over two terms as VP Projects.
Hands-on electrical engineering internship at an industrial electrical company in Calgary, Alberta. First industry exposure to hardware engineering right out of high school. Thank you to Justin and Curtis for their mentorship!
I'm always open to discussing new ideas! Whether you're working on AI, hardware, or the intersection of both — let's talk.
ericwen10@berkeley.edu →