AI/ML Engineer building models that go from notebook to production, not stopping at accuracy scores.
I work across the full ML lifecycle: data preprocessing, model training and tuning, explainability, and deployment into real systems, backed by a full-stack foundation that lets me ship the whole product, not just the model. This spans freelance client work, a mobile development role, and independent builds.
- 🔭 Currently building: multi-model ML systems with explainability baked in (SHAP, FinBERT), deployed via FastAPI
- 🌱 Deepening: LLM fine-tuning, RAG pipelines, agentic AI, MLOps
- 🧩 What separates my ML work: I don't hand off a model in a notebook. I build the APIs, databases, and interfaces that make it usable
- 💬 Ask me about: ML/DL model design, explainable AI, FastAPI backends, Flutter, full-stack AI products
- 📄 Published research: co-author, Journal of Applied Linguistics and TESOL (JALT), Vol. 9 (2026), DOI: 10.63878/jalt2260
|
Full-stack AI investment platform delivering explainable Buy/Hold/Sell signals for stocks & crypto. Ensemble: Random Forest · XGBoost · LSTM · GRU · Transformer Explainability: SHAP · FinBERT sentiment Stack: FastAPI · Celery · PostgreSQL · React/TS |
AI-powered lab report interpreter using OCR extraction, rule-based anomaly flagging, LLM-generated plain-English explanations, and trend tracking over time. Stack: FastAPI · PostgreSQL · React · Tesseract OCR 🔗 Live demo |
|
Detects waste objects from images and recommends recycle/compost/landfill disposal. Pipeline: YOLOv8 (detection) → EfficientNetB0 (classification) → XGBoost (recommendation) Stack: FastAPI · Flutter |
Interactive instrument for visualizing neural network internals live, with every activation, gradient, and feature map computed by a real PyTorch backend and nothing faked client-side. Stack: FastAPI · PyTorch · React · D3 |
