I'm Het Monpara, pursuing my M.Tech in Machine Learning at DA-IICT, Gandhinagar. My work centers on building intelligent systems end-to-end — not just training a model, but taking it from raw data through to something that's actually served, monitored, and used in the real world.
Currently building:
- Deep learning for wireless RF signal classification
- Sensor-based human activity recognition
- RAG-powered engineering assistants
- Cloud-native IoT monitoring and observability
- GPU demand forecasting and MLOps pipelines
- Multi-task ML for ride intelligence
Currently exploring: Vision Transformers, LLM agent systems, and distributed training
Deep Learning & Computer Vision
| Project | What It Does | Stack |
|---|---|---|
| RFVision | 21-class RF spectrogram classification — ResNet50, EfficientNet, ViT & CNN-ViT hybrids, LDAM + contrastive learning, Grad-CAM explainability | PyTorch TIMM Grad-CAM |
| GymSense AI | Hybrid CNN + Dilated Self-Attention on 7-channel sensor telemetry — activity recognition, rep counting, tempo scoring, LLM coaching | TensorFlow FastAPI React Groq |
Generative AI & Applied ML
| Project | What It Does | Stack |
|---|---|---|
| Engineering Intelligence Hub | Full-stack RAG platform — document ingestion, hybrid vector search, citation-backed LLM answers | FastAPI Next.js LangChain ChromaDB |
| Uber Ride Intelligence | Multi-task pipeline — ride-outcome prediction, CatBoost fare prediction, LSTM demand forecasting, K-Means segmentation | Scikit-Learn CatBoost Power BI |
| Shark Tank Intelligence | Three-task pipeline — deal prediction (62% acc.), valuation (R² 0.77), multi-label investor prediction | Pandas Gradient Boosting Lasso |
| STMP: Shark Tank Model Prediction | Earlier-stage model exploring deal-outcome prediction on Shark Tank pitch data — the groundwork that led to Shark Tank Intelligence | Python Scikit-Learn |
Infrastructure & MLOps
| Project | What It Does | Stack |
|---|---|---|
| GPU Cloud Forecasting | GPU demand forecasting with a full production ML lifecycle — tracking, serving, monitoring, dashboards | PyTorch XGBoost MLflow Grafana |
| Smart Irrigation System | Cloud-native IoT monitoring system for real-time sensor data — containerized services with managed observability on AWS | Python Docker Prometheus AWS Managed Grafana AWS SigV4 |
Web, Tools & Analysis
| Project | What It Does | Stack |
|---|---|---|
| DataDesk | Browser-based SQL learning platform — SQLite/WASM, Monaco Editor, ER diagrams | React WASM Supabase |
| frame2pdf | Web-based utility for converting image frames into a single PDF document | JavaScript |
| Cricket | Exploratory analysis and prediction modeling on cricket match data | Python Jupyter |
| IPL Win Predictor | Ball-by-ball IPL win probability prediction | XGBoost Streamlit |
| GujEstate AI | Gujarat real-estate valuation pipeline | Scikit-Learn Pandas |
| LinkedIn Recommendation | NLP-based professional recommendation system | Python NLP GenAI |
| MovieMingle | AI-powered movie discovery app | React OpenAI |
| Area | What I'm Exploring |
|---|---|
| Deep Learning | CNNs, Transformers, representation learning |
| Computer Vision | Image classification, transfer learning, Vision Transformers |
| Generative AI | LLM applications, retrieval-augmented generation, AI agents |
| Signal Intelligence | RF spectrogram classification |
| Infrastructure | Cloud-native monitoring, containerized ML services |
| Forecasting | Time-series modeling, resource and demand forecasting |
| Explainable AI | Grad-CAM and other model interpretation methods |
| MLOps | Experiment tracking, model serving, monitoring, deployment |
I care about the second half of that list as much as the first — a model that performs well in a notebook isn't the goal; a system that keeps performing well once it's serving real traffic is.
CNN ResNet EfficientNet Vision Transformers Transfer Learning LSTM XGBoost CatBoost Random Forest Contrastive Learning Grad-CAM RAG LLMs Vector Search Prompt Engineering
| Degree | Institution | Duration | Focus |
|---|---|---|---|
| M.Tech — Machine Learning | DA-IICT, Gandhinagar | 2025–2027 | CGPA: 8.31/10 |
| B.E. — Computer Engineering | GEC, Bhavnagar | 2021–2025 | CGPA: 8.61/10 |
