Marcel Ndowah

Marcel Ndowah ML/AI Engineer | Python · PyTorch · LangChain · FastAPI | Open to Remote Roles Buea, Southwest, Cameroon (CM) 500 connections • 1,406 followers About ML/AI Engineer with a Mathematics degree and a focus on building end-to-end machine learning systems - from raw data pipelines to deployed production APIs. My mathematics background (B.Sc., University of Buea) means I understand what happens inside ML models, not just how to call the library. Gradient descent, probability distributions, linear algebra - these are tools I think with, not concepts I look up. What I build: → Classical ML pipelines: data ingestion, feature engineering, model comparison, FastAPI serving → LLM applications: RAG systems, AI agents, tool-calling workflows → Full deployment: Docker, CI/CD, cloud hosting Open source contributions (all merged): → Hugging Face Transformers — Removed deprecated as_target_processor from Wav2Vec2 processor __call__ and pad methods, improving API clarity for downstream users → Hugging Face Transformers — Fixed chunked prefill implementation bug (issue 43082) → OCaml API Watch — Extended API diffing logic to detect class-level additions and removals Current stack: Python · scikit-learn · XGBoost · PyTorch · LangChain · FastAPI · Docker · PostgreSQL · ChromaDB Currently building production ML projects and open to remote ML/AI Engineer roles globally. ndowahmarcel@gmail.com github.com/marcndo Total Experience: 2 years and 6 months Experience Backend ML Engineer - AIMLSE - Adaptive AI Coding Platform(https://www.linkedin.com/company/aimlse) (Current) Apr 2026 - Present (2 months) in Remote - Resolved 3 critical bugs in a bidirectional block-to-Python code converter, raising test coverage from 27/30 to 30/30 and eliminating all known conversion failures - Restructured backend package architecture across 6 modules, resolving import conflicts and enabling clean integration between previously incompatible components - Identified and patc