Opeoluwa Osho

Details

Skills

Agentic AI & Multi-Agent Systems — LangGraph · CrewAI · MCP (Model Context Protocol) · event-driven multi-service AI architectures · autonomous agent pipeline design · tool-use and agent-action patterns · QA gating, exactly-once event processing, retries · agent orchestration at scale (Cloud Pub/Sub, Cloud Tasks)
LLM Evaluation & Observability — Automated QA gating before spend · eval harnesses
|
and grounding checks · production eval loops (live benchmarking → prompt selection) ·
|
citation/groundedness verification · LangSmith
Workflow Automation — n8n (incl. self-hosting & AI nodes) · Make.com · Zapier · LLM-to-SaaS integration
Backend & Full-Stack — Go / Golang · Python · TypeScript / JavaScript · Fastify · Prisma · Server-Sent Events (SSE) · microservices & event-driven architecture · RESTful & WebSocket APIs · React · Databases: PostgreSQL · MongoDB · MySQL · Redis · ElasticSearch
AI Infrastructure & Serving — GPU inference orchestration (RunPod, AWS SageMaker) · vector databases (Pinecone, pgvector) · dynamic request batching · concurrent job queues & scheduling · inference latency optimization
Infrastructure & Delivery — Docker · Kubernetes · Linux · Terraform · CI/CD (GitLab, CircleCI, Jenkins) · AWS · Google Cloud Platform (Cloud Run, Pub/Sub, Cloud Tasks, Secret Manager) · mTLS / secure communication

About

AI and backend engineer with 7+ years of production backend experience across Go, Python, and TypeScript, now building AI systems end to end — an AI advertising platform I architected from scratch on Google Cloud (exactly-once event processing, command queues, automated QA gates), sustaining 20–30k ads/day; RAG document intelligence for finance teams; and Stable Diffusion inference serving 2–4M monthly visits. I bridge the gap between LLM APIs and real business systems: agentic orchestration, retrieval pipelines, and event-driven services that stay correct under load, retries, and partial failure.