All work

Collinson Group2023 — Present

Unified Data & AI Platform

One platform for the full ML lifecycle — experimentation, training, deployment, real-time inference, CI/CD, observability and governance — shared across teams.

  • ~60%faster deployment cycles
  • 1platform shared across teams

Architecture · simplified

  1. DataSnowflake as the source of truth
  2. Experiment & trainReproducible, tracked runs
  3. OrchestratePrefect pipelines
  4. Deploy & serveCI/CD, real-time inference on AWS
  5. Observe & governMonitoring, lineage, controls

Context

As the number of models and teams grew, each project was rebuilding the same plumbing — its own pipelines, its own deployment scripts, its own idea of monitoring. Shipping was slow and running things in production was fragile.

The challenge

Build a shared foundation that data scientists actually want to use, without slowing them down — and that engineering, security and governance teams can trust.

What I did

  • Designed and built a unified Data & AI platform on Snowflake, AWS and Prefect, covering experimentation, training, deployment, real-time inference, CI/CD, observability and governance.
  • Established production engineering practices — CI/CD for models, observability, model lifecycle management and reusable infrastructure — as defaults rather than afterthoughts.
  • Led the data science and ML engineering team that built and ran it, and set the standards teams adopted.

Outcome

Deployment cycles dropped by ~60%, reliability improved, and new use cases start from a paved road instead of a blank page.

↑↓ move↵ open