Spent 6 years as Staff Engineer on Meta's core ML infrastructure team, scaling model serving from millions to billions of daily inferences. I specialise in production ML system design, model optimisation, and helping engineering teams bridge the gap between research prototypes and reliable production systems. If your model works in a notebook but not in prod, let's talk.
4.9
9 reviews
James T.
✓ VerifiedHe reviewed our model serving setup in 20 minutes and found 3 bottlenecks we'd been chasing for months. Unreal depth of knowledge — this was worth 10x the price.
Response from Marcus Obi
The batching issue was a classic one — glad we caught it before your scale-up.
Priya M.
✓ VerifiedMarcus saved us from a very expensive architecture mistake. His knowledge of model serving tradeoffs is unparalleled — he's clearly seen every failure mode.
Sofia L.
✓ VerifiedMarcus has the rare ability to explain complex ML infrastructure tradeoffs in plain English. Our entire eng team is now aligned on the architecture thanks to his framing.
Ben C.
✓ VerifiedVery strong on PyTorch specifics. I had hoped for more discussion of cloud infrastructure but Marcus was happy to pivot once I asked. Really solid session.
Rachel W.
✓ VerifiedHis tip on batching strategy alone cut our inference latency by 40%. That's a session that paid for itself in the first week.
