Skip to content
View Eatosin's full-sized avatar
💭
building...
💭
building...

Block or report Eatosin

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
eatosin/README.md
Owadokun Tosin Tobi

Senior AI Engineer — autonomous agents, LLM evaluation, developer tooling

Lagos, Nigeria · Open to remote

LinkedIn Email Hugging Face

I build agents that survive production: evidence-gated RAG, evaluation harnesses that gate every release, and tenant-isolated deployments where access control is real. Three years of shipping it — mostly Python and TypeScript on managed GCP.

Right now

  • 🔧 Shipping dshify — an open-source CLI + DeepSeek Harness plugin that ports agent-harness setups (OpenCode first) to DSH with verification. ~1.8k weekly npm downloads, SLSA provenance, 171 tests.
  • 🔒 Contracting on privacy-preserving enterprise RAG inside agency-managed GCP: multi-tenant isolation, clearance-level access control, and an eval suite shipped in the same release.
  • 🎓 Teaching AI engineering at Torilo Academy — 10+ interns trained and certified on core SWE and applied ML.
  • ✍️ Writing verified teardowns of agent security and hallucination failures.

Featured work

Project What it does Signal
dshify CLI + DSH plugin for agent-harness migration, with verification ~1.8k weekly npm downloads · 171 tests · SLSA provenance
Axiom Engine Autonomous auditing OS + MCP server — LangGraph, FastAPI, Next.js 99%+ answer accuracy · sub-4s p95
DocuTrace AI Auditor Verifiable document extraction on Google LangExtract + Gemini source-grounded citations
Sentinel Anomaly detection + RAG root-cause agent >2.5σ drift thresholds · live API
NewsAgent Pro Multi-agent research → write → critique pipeline 60s prompt-to-post
Spectre Physics-informed deepfake audio detection ResNet over Mel-spectrograms

Open source

google/langextract — shipped the thinking_config allowlist fix. Gemini 3 / thinking models were having the parameter silently filtered out, which forced the default high thinking budget and caused latency and timeouts (issue #319). Landed upstream through PR #553, credited as co-author in the squash commit.

Writing

Toolbox

Python, PyTorch, TypeScript, Next.js, FastAPI, Docker, Postgres, GCP, Redis

LangGraph · LangSmith · MCP · RAGAS · OpenCode · Pydantic · Supabase · pgvector · CI/CD

Pinned Loading

  1. NewsAgent-Pro NewsAgent-Pro Public

    Autonomous AI Newsroom that researches, writes, and designs content.

    Python 6

  2. Zeta Zeta Public

    Zeta: Physics-Inspired AI Test Generator — Gemini 2.5 + Z-Score/Isolation Forest for risk-aware, edge-case-focused automated QA. FastAPI backend — generate executable Selenium tests instantly.

    Python 2

  3. Sentinel-MLOps Sentinel-MLOps Public

    Autonomous Anomaly Detection & RAG Investigation Agent (FastAPI + Gemini). Use this for your project website.

    Python 2

  4. PolyMind-Crypto-Arbitrage PolyMind-Crypto-Arbitrage Public

    AI-Gated Arbitrage Strategy for Prediction Markets. Uses XGBoost to predict trade fill probability and avoid liquidity traps in volatile order books.

    Jupyter Notebook 5 1

  5. TalentAlign-AI-Recruiter TalentAlign-AI-Recruiter Public

    Unbiased automated LLM-powered resume screening app solving bias, volume, and fit challenges.

    Python 2

  6. Spectre-Deepfake-Detection Spectre-Deepfake-Detection Public

    Physics-Informed Deepfake Audio Detection using Spectrograms and CNNs.

    Jupyter Notebook 3