I'm an AI/ML Engineering student at SRM Institute of Science and Technology (B.Tech CSE — AI & ML), building systems that go beyond notebooks — production APIs, experiment-tracked pipelines, LLM-powered applications, and RAG architectures.
My work lives at the intersection of machine learning engineering and software engineering: I care equally about model quality and the infrastructure that makes it deployable, observable, and maintainable.
| Domain | What I Build |
|---|---|
| LLM Applications | Retrieval-augmented generation, document Q&A, prompt pipelines |
| LLM Internals | GPT architecture from scratch, attention mechanisms, tokenization |
| ML Engineering | End-to-end pipelines from raw data to trained model to monitored API |
| MLOps | MLflow experiment tracking, Docker packaging, FastAPI deployment |
| Churn & Risk Modelling | XGBoost with RFM segmentation, threshold calibration, SHAP explainability |
| Fraud Detection | Ensemble ML, anomaly detection, real-time scoring APIs |
End-to-end churn prediction pipeline for a direct-to-consumer e-commerce business.
Stack: XGBoost · FastAPI · MLflow · Docker · Python
| Part | Focus |
|---|---|
| Part 1 | EDA, data quality audit, churn-risk hypothesis formation |
| Part 2 | RFM feature engineering, cohort segmentation |
| Part 3 | XGBoost training, MLflow experiment tracking, threshold calibration |
| Part 4 | FastAPI deployment, Docker, CI/CD, monitoring plan |
Multi-PDF conversational AI using LangChain and FAISS vector embeddings.
Stack: Python · LangChain · FAISS · Streamlit · OpenAI · Repo
Real-time fraud detection with ensemble ML, SHAP explainability, and FastAPI serving.
Stack: Python · XGBoost · SHAP · FastAPI · Docker · Repo
ML pipeline on telecom churn with feature engineering, model comparison, and SHAP interpretation.
Stack: Python · Scikit-learn · Pandas · Matplotlib · Repo
Production MLOps template: experiment tracking, model registry, CI/CD, and drift monitoring.
Stack: Python · MLflow · DVC · FastAPI · Docker · GitHub Actions · Repo
Studying and extending Karpathy's minimal GPT-2 implementation. Added fine-tuning guide and benchmark notes.
Stack: Python · PyTorch · Transformer Architecture · Repo
Working through structured ML engineering curriculum. Logging experiment results and key production patterns.
Stack: Python · MLflow · FastAPI · Testing · Repo
Exploring prompt flow DAGs, evaluation frameworks, and variant A/B testing for production LLM apps.
Stack: Python · Promptflow · LLM APIs · RAG · Repo
LLM-powered legal document Q&A using retrieval-augmented generation.
Stack: Python · LLM API · RAG · Vector Embeddings
SIH 2025 finalist. Multilingual AI platform for environmental education.
Stack: React · FastAPI · OpenAI/Gemini · Multilingual NLP
- RAG architecture patterns — dense retrieval, re-ranking, hybrid search
- LLM internals — GPT architecture, attention, fine-tuning with LoRA/QLoRA
- MLOps depth — model drift detection, data pipeline observability
- Agent frameworks — LangChain, Promptflow, tool-use orchestration