Great on Count Subarrays with Sum Equal to K, Difficulty Easy
Great on Classification Model Inaccuracy, Difficulty Easy
Great on Diagnose and Fix RAG Pipeline Retrieval Issues, Difficulty Medium
Built and tuned 6+ ML models (Regression, Classification, CNNs) using SMOTE and hyperparameter optimization, improving accuracy by 10–25%.
Designed reproducible ML pipelines for preprocessing and feature engineering, integrating MLflow for experiment tracking and versioning.
Improved model performance by up to 30% using SVD and Transfer Learning.
Developed a Retrieval Augmented Generation (RAG) system using LangChain and ChromaDB to enable semantic search over unstructured corporate data. Orchestrated data versioning pipelines using DVC synced with AWS S3, and implemented MLflow to track prompt engineering experiments and embedding performance. Built a fully automated deployment pipeline using GitHub Actions and Docker; commits trigger automated Pytest suites and seamless deployment to AWS EC2. Engineered high-performance Async APIs using FastAPI to handle concurrent document ingestion, chunking, and vector retrieval.
Architected a stateful conversational agent using Dialogflow ES and FastAPI, handling complex transactional workflows (ordering, modifications). Migrated persistence layers to a production-grade MySQL database on AWS RDS, managing schema design and session integrity. Deployed infrastructure on AWS EC2 (Linux), managing persistent background processes using Screen and implementing Ngrok tunneling for secure SSL communication.