Marcus Oduya

Marcus Oduya is a practitioner's practitioner. Over the past eight years, he's led AI implementation programs at organizations ranging from 200-person startups to global enterprises with tens of thousands of employees — and he's done it in environments where failure has real consequences.

His expertise is production reliability: the specific, unglamorous work of making agents that demo beautifully actually hold up when they encounter real users, real data, and real failure conditions.

Marcus is particularly focused on:

• Agent orchestration patterns that degrade gracefully under load
• Human-in-the-loop design — knowing when to hand back to a human and doing it without breaking the user's trust
• Failure taxonomy for business-process agents: the patterns that kill projects in their first 90 days

He's spoken at re:Invent, Databricks Data + AI Summit, and numerous internal engineering conferences where he can actually be candid about what goes wrong.

"Every agent failure I've ever investigated had a simpler root cause than the team expected. Usually it was an assumption baked into the design that nobody thought to question."

Marcus is based in Lagos and London and is available for select advisory engagements.