Projects
Explore my latest work in AI and Data Engineering
University Q&A System with LLM & RAG
Built an LLM-powered Retrieval-Augmented Generation (RAG) system that allows users to ask natural language questions about 100+ universities. Integrated data from Quora and Reddit using scalable crawlers. Designed chunking, metadata tagging, and vector search using LangChain and OpenAI to return accurate, context-aware responses.
Search Query Trend Prediction Platform
Engineered a data pipeline to process 100M+ monthly Google SERP records and forecast keyword trends using clustering and regression models. Handled large-scale data ingestion, transformation, and analytics for business insights.
ML-Based Topic Categorization Engine
Created an end-to-end machine learning pipeline to classify search queries into relevant marketing categories. Implemented text preprocessing, model training using fine-tuned BERT, and deployed real-time prediction APIs on Google Cloud. Used for search analytics and content tagging.
LLM-Powered Resume Builder using Agents
Developed a multi-agent LLM system to generate customized candidate resumes based on job descriptions and career goals. Orchestrated different agents for formatting, tone control, and ATS optimization using LangChain.
LLM Fine-Tuning for Tech Store Inventory Support
Fine-tuned a lightweight local LLM to support inventory-based Q&A for a tech store. Enabled natural language queries about stock levels, pricing, accessories, and restocking via a chatbot interface. Focused on domain-specific vocabulary and retrieval consistency.
Kafka-Based Warehouse Notification System
Designed a real-time Kafka-based notification system to monitor warehouse inventory updates and send alerts across services. Integrated multiple data sources using producers and consumers with fault-tolerant event handling.
Scalable Web Crawlers for Business Intelligence
Built robust, multi-threaded crawlers for platforms like Quora, Reddit, LinkedIn, and Twitter. Collected millions of records with error handling, deduplication, and real-time storage into MongoDB. These datasets fuel various NLP and analytics pipelines.
QA System for CSV & Excel Files (LLM-Based)
Built a question-answering system that enables users to ask questions about large structured datasets (CSV/XLSX) without using RAG. Handled preprocessing, indexing, and dynamic summarization using OpenAI models with context window optimizations.
LLM-Based Stock Sentiment Analyzer
Developed a sentiment analysis tool for stock-related discussions using LLMs like GPT. Extracted and analyzed user emotions from Reddit and Twitter using prompt-engineered queries. Translated sentiment trends into actionable investment signals for analysts.