Silicon Engineering Platform

A post-silicon validation analytics platform: it generates realistic synthetic silicon test data, parses and decodes it with a C++ core (exposed to Python via pybind11), stores it in a relational database, runs yield & failure analysis, and serves it through a FastAPI + React dashboard with Schmoo plots and wafer maps — plus automated HTML/PDF reports.

Built to demonstrate post-silicon validation domain skills alongside production-grade software engineering across C++, Python, SQL, and web. See BUILD_PLAN.md for the original (historical) design roadmap, and CONTRIBUTING.md for current setup/learning.

Dashboard

Yield summary, failure Pareto, wafer map (spatial pass/fail with realistic radial yield loss), and a Schmoo plot (voltage × frequency operating envelope). The core: cpp badge indicates the compiled C++ parser is active.

Auto-generated REST API (FastAPI / OpenAPI):

Architecture

 datagen (Python) ──► tests.log / *.csv ──► C++ core (parse + decode, pybind11)
        │                                          │
        └──────────────► ingest (SQLAlchemy) ──► SQLite ──► analysis (pandas)
                                                     │            │
                                            FastAPI API ◄─────────┘
                                                     │
                                          React dashboard + reports

Quickstart

Option A — Docker (one command)

docker compose up --build
# API:       http://localhost:8000/docs
# Dashboard: http://localhost:5173

Option B — Local

Requires Python 3.10+, a C++17 compiler, and CMake (for the compiled core; without it the platform automatically falls back to a pure-Python parser).

# 1. Install (compiles the C++ extension via scikit-build-core + pybind11)
pip install -e ".[dev]"

# 2. Generate data, load the DB, and write a report in one shot
sep demo

# 3. Serve the API, then run the dashboard
sep serve                     # http://localhost:8000
cd frontend && npm install && npm run dev   # http://localhost:5173

Check which parsing backend is active:

sep info        # -> core backend: cpp   (or: python)

CLI

Command	Purpose
sep generate	Generate a synthetic post-silicon lot
sep ingest <lot_dir>	Load a lot into the database
sep yields	Print the yield summary
sep report <lot_id>	Write an HTML report (--pdf for PDF)
sep decode 0xA000830D	Decode a raw register value into fields
sep serve	Run the FastAPI server
sep demo	generate → ingest → report in one step

API endpoints

/api/health, /api/yield, /api/pareto, /api/wafers, /api/wafermap/{n}, /api/schmoo/{die_id}, /api/dies, /api/registers/{die_id}. Interactive docs at /docs.

The C++ core

Two compiled components live in cpp/ and are exposed as the sep_core module:

• log_parser — a streaming, zero-copy (string_view) test-log parser.
• reg_decoder — spec-driven bitfield extraction and expected/actual compare.

They sit on the critical data path (used by ingest and the API). If the extension isn't built, sep/core.py selects the pure-Python fallback in sep/_pyfallback.py (identical interface).

Testing

pytest            # Python: core + end-to-end pipeline

Contributing / learning the codebase

See CONTRIBUTING.md for dev setup, common commands, and a guided code-reading tour (a fast path to understanding the C++/pybind11 core, the SQL layer, and the full-stack flow).

Project layout

cpp/        C++ core (log parser, register decoder) + pybind11 bindings
sep/        Python package: datagen, db, analysis, api, reports, cli
frontend/   React + Vite dashboard (Plotly)
specs/      Example register specification
tests/      pytest suite

License

MIT (see the repository for details).