Hello there π, I'm Emmanuel Samuel
I'm a software engineer who works mostly on AI systems. Recently I built LangGraph agents that serve seven product lines from one platform, and an AIOps console that pulls six monitoring systems into one place and ranks the likely cause when something breaks.
My portfolio is at emmanuelmayowasamuel.vercel.app. It has an AI Twin you can chat with about my work.
I also keep a hand in classical ML. My sports predictor combines Poisson regression, XGBoost and Elo ratings.
I'm open to remote AI/ML Engineer, Forward Deployed Engineer and Software Engineer roles (US/EU).
- Multi-Tenant Agentic AI Platform. LangGraph agents on AWS Bedrock AgentCore serving 7 product lines, loading their tools at runtime over MCP.
- Enterprise RAG Platform. Retrieval on AWS S3 Vectors with Bedrock Titan v2 embeddings, one vector index per tenant.
- AIOps Operator Console. Six monitoring systems behind one connector SDK, with anomaly detection and root-cause ranking on top.
- Per-Call LLM Cost Instrumentation. Prices prompt, output, thinking and image tokens separately, per call, against a per-model rate table.
devdotfast/whiteboard Β· open-source canvas for thoughtful software design Β· β 3.2k
| Issue | Fix | What it solved |
|---|---|---|
| #1222 | #1223 β merged | Desktop hung at start-up on Windows because it sent SIGUSR2, which Windows doesn't support. It now checks the old server is really the same process, then stops it with SIGTERM. |
| #1224 | #1225 β merged | Claude Code (claude.exe) wasn't detected on Windows. Agent CLIs are now found by their .exe and .cmd names. |
Benaiah Studio Β· founder
- figma-to-code: a free, read-only bridge from Figma to code that works on Figma's free plan. It exports frames to files, writes a compact spec and design tokens, and checks the built page against the design. v0.1 beta, MIT.
- figma-bench: an open, reproducible benchmark for Figma-to-code tools, run on CC BY 4.0 Figma Community frames. v0.1 beta, round 1 results published. MIT.



