Building production-grade business systems for retail, FMCG and distribution —
where the hard part is never the code, it's the process behind it.
- 🛠️ Currently building an HRIS with face recognition (InsightFace + ONNXRuntime, decoupled
face_serviceover HTTP). - 💻 Most days I'm deep in Python / FastAPI / MySQL / PostgreSQL / Node.js / Docker.
- 📊 I design ERP integrations, pricing governance and inventory intelligence at 80,000+ SKU scale.
- 👁️ I ship computer vision to the warehouse floor: YOLO segmentation + OpenCV + ArUco for automated CBM measurement.
- 📈 Currently exploring ML-driven signal generation for Sharia-compliant IDX equities.
- ⚡ Fun fact: 10 years in, and I still start every system by mapping the business process, not the schema.
People Counting Engine — real-time occupancy counting from live CCTV feeds. TensorRT inference over RTSP streams, tuned across a GPU migration (GTX 1650 → RTX 5060 / Blackwell sm_120) on a CUDA 12.8 · TRT 10.8 stack. Production-deployed.
CBM Measurement Assistant — automated carton dimensioning from a single photo. YOLO segmentation, ArUco-based calibration and perspective correction on a controlled mat. Replaces manual tape-measure workflows in inbound logistics.
HRIS + Face Service — attendance and identity verification built on ArcFace 512-dim embeddings, isolated behind an HTTP boundary so the ML runtime never blocks the app tier.
IDX Signal Engine — an eight-layer research stack (ingestion → Sharia screening against OJK DES → feature engine → strategy → backtesting → risk → execution adapter → monitoring).
More in the pipeline: pricing governance tooling and product intelligence write-ups.
Systems outlive code. Build for the process.