I build end-to-end machine learning systems — from data pipelines and model training to production APIs and dashboards. My work spans predictive maintenance for industrial machinery, computer-vision quality inspection, RAG/LLM applications, and MLOps. I care about code that actually ships: tested, containerized, and reproducible.
- Focused on ML engineering, applied AI, and data infrastructure
- Real-world experience with industrial / manufacturing analytics (CNC machines, robotics, defect detection)
- I ship projects with tests, CI, and Docker — not just notebooks
- Reach me: harshalingawale48@gmail.com
Languages
ML / AI
Data & Backend
| Project | What it does | Stack |
|---|---|---|
| Predictive Maintenance Dashboard | Real-time machine health scoring, ML failure prediction (RUL + anomaly detection), bilingual dashboard, automated shift-handover PDF reports. | Dash · scikit-learn · lifelines |
| Industrial Defect Detection | Visual quality inspection with two approaches — deep anomaly detection (PatchCore + FAISS) and classical SSIM. | PyTorch · FAISS · OpenCV |
| LLM RAG Assistant | Retrieval-augmented Q&A service over your own docs. Runs offline; plugs into an LLM with one env var. | FastAPI · sentence-transformers · FAISS |
| MLOps Model Serving | Train → version → serve → containerize → CI. A model-quality gate runs on every push. | scikit-learn · FastAPI · Docker · Actions |
| Data Pipeline (ETL) | Modular Extract→Transform→Load with a data-quality gate that blocks bad data. | pandas · SQLite · pytest |
| Customer Churn Analysis | Full DS study: EDA, interpretable model (AUC ≈ 0.84), and retention recommendations. | scikit-learn · Jupyter · matplotlib |
