My projects across different technologies and domains.
An end-to-end enterprise platform that predicts customer churn, validates retention campaigns through A/B testing, and uses Agentic AI/RAG to generate personalized re-engagement offers.
A deterministic validation layer that catches numeric hallucinations in LLM output by checking extracted numbers, percentages, currencies and ratios against ground truth. No LLM in the loop — every verdict is rule-based and reproducible.
A self-hosted, multi-agent knowledge operating system built on Open Notebook, adding agent orchestration, verified research, and study-material generation on top of a privacy-focused research assistant.
Multi-agent version of the portfolio assistant with advanced UI/UX, charts and financial metrics. Coordinates Controller, Planning, Reflection, Critic, Decision and other agents to turn market data, news and RAG into a single investment strategy.
A single-agent LLM chat assistant for portfolio Q&A. Built with LangGraph, RAG and ChromaDB; fetches live stock data via an MCP stock-market API.
A data analytics project that segments customers using demographics, web/app usage and purchase history. Reached 97.38% accuracy with an optimized XGBoost model.