All work
MLOps & Data · DARVIS · 2020 — 2023
400+ Model Training Factory
Automated end-to-end training, conversion and deployment that shipped 400+ production vision models.
01 The problem
Every client needed models retrained on their own data — manual training could not keep up.
02 What I did
- Automated training and model conversion in Python with NVIDIA TAO, integrated with AWS S3.
- Tracked experiments with Weights & Biases, ClearML and DVC; deployed in Docker on cloud.
03 Impact
- 400+ models trained and deployed
- Repeatable, tracked experiments
Client work — details generalised to respect confidentiality. Happy to discuss specifics in an interview.