2026
in-progress
Featured
Agentic EDA Pipeline (DRDO SSPL)
An agentic AI pipeline that turns natural-language circuit specifications into simulated, layout-ready designs.
- Year
- 2026
- Category
- Agents
- Stack
- Python, LLM agents, RAG, Design-rule checking
- Role
- Team lead, six interns
Summary
Built at the Solid State Physics Laboratory (SSPL), DRDO. Covers schematic capture, datasheet retrieval (RAG), layout generation and design-rule validation.
What it does
A user describes a circuit in plain language. The pipeline captures a schematic, retrieves component data from datasheets with RAG, generates a layout and validates it against design rules. A working prototype exists and the team is adding capabilities to meet the expectations of the lab's potential users.
Design decisions
- Pass/fail decisions sit in deterministic checks, not in the LLM.
- Computational resource use is tracked, and the system is designed for local, on-premises deployment under the lab's security requirements.
- I made the major architecture and technical decisions, worked hands-on across the RAG pipeline and design-rule verification, and wrote the pipeline's test harness.