Priyansh Patel

AI Engineer · Production Systems · Surat, India

Building AI systems that work outside the lab.

I design and ship computer vision, deep learning, and offline-first software for real environments. Recent work includes a live retail loss-prevention platform and a production garage management system.

  • MSc AIDistinction
  • 0.941Chitr ROC-AUC
  • Offline-firstJalaSai in production

Snapshot

Professional summary

What I build

Computer vision systems, active learning loops, and practical software that balances latency, reliability, and operational simplicity.

How I work

I start from the deployment reality, then shape the model, backend, and UI around the actual user constraints.

What I value

Measurable outcomes, clear trade-offs, and systems that keep working when the happy path breaks.

Selected work

Projects that show production judgment

Live deployment

Chitr - Real-Time Loss Prevention

Retail theft detection pipeline that uses YOLOv8 as a fast gate and VideoMAE / SlowFast R101 for asynchronous clip analysis. PostgreSQL feedback and MLflow tracking keep the system improving in the field.

  • 0.941 ROC-AUC
  • 93.3% accuracy
  • 2.7% false alarm rate
PyTorchVideoMAEYOLOv8FastAPIPostgreSQLMLflow

Journey

Experience and education

Skills

Tools and domains I use often

Contact

Open to AI engineering, applied ML, and production system roles.