I'm an AI and machine learning engineer with a strong focus on healthcare AI and data engineering. My work spans deep learning for medical imaging, clinical data preprocessing at scale, and building interactive tools for understanding neural networks. I'm proficient in both backend and frontend development, with experience building production-ready systems that process complex datasets and make machine learning interpretable.
| Area | Expertise | Evidence |
|---|---|---|
| Medical AI | Brain tumor classification with Grad-CAM | ~92% accuracy, explainable predictions |
| Clinical Data | MIMIC-IV preprocessing pipeline | 40GB+ dataset, production-grade |
| Deep Learning | Neural network visualization & education | 255+ tests, real PyTorch hooks |
| Full-Stack | React + FastAPI/Express systems | Multiple deployed applications |
| Mobile | Flutter cross-platform apps | Firebase, real-time features |
A production-grade educational platform for understanding neural networks from first principles
Live Instruments:
- ✅ Network Canvas - Semantic-zoom D3 graph with VCR-style stepping
- ✅ Activation Lab - Interactive curves & freehand drawing
- ✅ CNN Lab - Filter factory, 3D features, receptive fields, saliency
- ✅ Optimizer Arena - Loss surfaces & optimizer racing
- ✅ BatchNorm Tracker - Real-time feature normalization
Why Unique: 255 tests, zero faked data. Every number comes live from PyTorch via hooks.
Healthcare AI that explains its decisions in real-time
Key Metrics:
- 🎯 91-92% accuracy on 4-class classification
- 📊 Grad-CAM explainability for model interpretability
- 🎨 Interactive Streamlit dashboard
⚠️ Transparent limitations documented
Enterprise-grade preprocessing for 40GB+ clinical datasets
5-Stage Pipeline:
- Load (schema inference) → 2. Clean (validation) → 3. EDA (analysis) → 4. Features (engineering) → 5. Datasets (outputs)
Output: 6 ML-ready Parquet datasets for prediction, embeddings, and NLP tasks
Distributed big data processing for large-scale analytics
Analytics: Revenue analysis · Customer metrics · Spend classification · Rolling averages
Features: JWT auth · Rate limiting · Fee reconciliation · S3 uploads · PDF generation
Stack: Frontend (React + Vite) · Backend (Express) · Database (MongoDB) · Cloudinary uploads
Features: Real-time chat · AI translation · Media sharing · Rive animations
Features: Real-time location tracking · Provider matching · In-app notifications
| Category | Focus Areas |
|---|---|
| 🏥 Healthcare AI | Medical imaging, Clinical data pipelines, AI explainability |
| 🧠 Deep Learning | Interactive visualization, Theory meets practice, Neural network education |
| 💾 Data Engineering | MIMIC-IV preprocessing, Distributed processing, ML-ready pipelines |
| 🔄 Full-Stack | PyTorch → FastAPI → React, Production systems, User-centric design |
AI/ML Development ████████████████░░ 90%
Data Engineering ████████████░░░░░░ 80%
Backend Development ██████████████░░░░ 85%
Frontend Development ███████████░░░░░░░ 75%
DevOps & Deployment ██████░░░░░░░░░░░░ 40%
Research & Interests:
- 🏥 Healthcare AI & Medical Imaging
- 🧠 Explainable AI (XAI) & Interpretability
- 📊 Large-scale Data Processing
- 🎓 ML Education & Visualization
- 🚀 Production ML Systems
