System online · building since 2022
AI & MLOps Engineer and Backend Developer. I take machine learning out of the notebook and into production — training pipelines, containerised services, and the APIs that hold them together.
01 · Profile
Backend fundamentals, applied to machine learning. I care about what happens after the model works — validation, deployment, observability, and what it costs to run.
I'm Anthony Njoku, an AI, MLOps and Backend Engineer who builds systems that solve real problems rather than demos that stay in a notebook.
My work centres on AI and MLOps pipelines — integrating machine learning models, automating the path to production, and keeping deployments observable and scalable. I write services in Python (FastAPI, Flask) and Node.js (Express), and I've shipped larger platforms in PHP & Laravel.
On the data side I work with MySQL and PostgreSQL; on the infrastructure side with Docker, Kubernetes, MLflow and CI/CD on AWS and Azure. That covers API design, automated training runs, real-time analytics, and monitoring of models once they're live.
02 · Selected work
Production services, automation tools and a few experiments. Filter by the part of the stack you care about.
9 projects
A production RAG service over your own documents. Hybrid retrieval (vector + BM25) with cross-encoder re-ranking, and a strict abstention path: if the answer isn't in the uploads it says so instead of guessing. Deployed on infrastructure I run myself — Docker Compose behind Nginx on Oracle Cloud, with Cloudflare handling DNS, CDN and TLS for the custom domain. A 30-question golden set scores correctness, faithfulness and abstain accuracy on every change.
A Rust service backing a Telegram bot: an AI analyst for SOL, BTC, ETH and BNB that reads Bybit perpetuals across three timeframes — funding, open interest, order-book depth, VWAP, EMA200 and clustered support and resistance with touch counts. Signals must pass hard risk rules or they are discarded, and every one is tracked to target or stop with the hit rate published. It never places an order.
A trained model served behind FastAPI with automatic input validation, real-time predictions, and interactive documentation through Swagger UI.
A command line you can talk to. Natural language is transcribed, resolved by an LLM and executed in real time, with spoken feedback — speech-to-text, generation and text-to-speech in one loop.
// repo link pending
A Gemini-powered assistant supporting both voice and text, built for hands-free conversation and reusable across conversational products.
// repo link pending
A working virtual top-up platform for airtime, data and utility bills — third-party API integration, secure payment processing, and an end-to-end transaction flow.
A RESTful task service in Node.js and Express — clean, scalable endpoints, authenticated access, and a documented data flow between client and server.
// repo link pending
Scan and connect — no password typing for guests, customers or staff, and no handing out network credentials to everyone who visits.
A trading education platform with structured course material, mentor access and live market insight for members.
No projects match that filter yet.
03 · Contact
If you need an API, a model in production, or a manual process automated end to end, send me the problem and I'll tell you how I'd approach it.