All Projects
Open-source projects spanning LLMs, ML, and data engineering
AI Research Agent
Autonomous research agent that decomposes queries, executes multi-turn tool calling with web search and arXiv, and validates completeness through self-reflection. Features real-time Streamlit UI.
Weaviate MCP Inspector
Natural language interface for Weaviate vector databases through Claude using Model Context Protocol. Enables intuitive database exploration via conversation with 9 inspection tools.
Handwritten Text Recognizer
Deep learning OCR system with ResNet encoder and Transformer decoder (14M parameters). Achieves 70% error reduction via augmentation. Deployed as FastAPI microservice on GCP with monitoring.
Movie Recommendation System
Netflix-inspired collaborative filtering system using matrix factorization (SVD, SVD++) on 480K+ users and 17K+ movies. Implements KNN-based recommendations with comprehensive feature engineering.
Social Network Link Prediction
Graph ML solution for predicting missing links in Facebook social network (1.86M nodes, 9.4M edges). Engineers 30+ graph features achieving 0.92 F1 score with Random Forest.
Tweet Sentiment Extractor
Transformer-based NLP model that extracts sentiment-bearing phrases from tweets. Fine-tuned question-answering implementation for Kaggle competition with comprehensive error analysis.
ETL Pipeline
End-to-end data pipeline orchestrating NYC taxi data from extraction to BigQuery loading. Uses Apache Spark on Dataproc with Prefect orchestration and Terraform for infrastructure as code.
Showing 7 projects