$ Building intelligent systems · ML Research → Production
I build ML systems end-to-end — from algorithm design to production-ready pipelines. My work spans Computer Vision, NLP, RAG architectures, and deep learning for real-world IoT and smart-home systems.
- Implementing ML algorithms from scratch to master the fundamentals
- Designing production-grade RAG systems with hybrid retrieval & re-ranking
- Researching multi-task deep learning for smart-home activity prediction
- Applying transfer learning to industrial fault diagnosis under real-world conditions
| Project | Description | Stack |
|---|---|---|
| Bearing-Fault-Diagnosis | Comparative study on vibration signal representations for industrial fault diagnosis using transfer learning and domain adaptation for vibration-based predictive maintenance | PyTorch Transfer Learning Signal Processing |
| Ask-My-Documents | Production-grade RAG system with hybrid retrieval, re-ranking, and source-grounded document QA across uploaded PDFs and documents | RAG LangChain LLM Python |
| Predictive Human Activity Modeling | Achieved 98% activity accuracy and 0.93 macro F1 using a hybrid GAT + BiLSTM architecture with activity-conditioned time prediction heads. Predicted time-to-next-activity with 10–13 min MAE using quantile-based discrete time binning for skewed smart-home sensor intervals | PyTorch GAT BiLSTM PyG Multi-Task Learning |
| Legal-Document-Summarizer | Architected BART-based NLP summarization pipeline processing 50-page PDFs in under 10 seconds using sliding-window chunking for token-limit handling. Built FastAPI REST API supporting PDF, DOCX, and plain text inference | Python Hugging Face FastAPI PyPDF2 Transformers |
| IoT-Biometric-Attendance-System | Deployed IoT biometric attendance platform for 100+ users with MQTT-based AWS IoT Core synchronization at sub-2s latency and real-time monitoring dashboard | ESP8266 AWS IoT Core JavaScript MQTT |
| Movie-Recommender-System | Content-based movie recommendation engine using vectorized metadata and cosine similarity | Python Scikit-Learn Streamlit |
--
| Domain | Technologies |
|---|---|
| Languages |    |
| ML / DL |     |
| GenAI & NLP |  RAG Pipelines LLMs Vector DBs |
| Tools |     |
⌘ Built with curiosity & coffee · Anagha R S