Designed an end-to-end data pipeline for EEG and ECG analysis, including preprocessing of biomedical signal images, extraction of clean time-series data, and predictive modeling for physiological pattern detection. The project focuses on transforming raw biosignal data into reliable features for downstream ML prediction tasks.

Built an end-to-end AI agent with LangGraph that connects to databases, analyzes business documents, maps KPIs to industrial use cases, and generates ML pipelines for predictive insights and optimization.

Built an AI voice agent that autonomously calls moving companies to collect price estimates, using STT, TTS, and LangGraph-driven conversation flows. Delivered an MVP with real-time company search, automated call handling, and a Streamlit interface for live demos.

Developed a real-time data streaming pipeline with Kafka and Spark, replacing a legacy batch ETL system. Delivered ML-powered dashboards for predictive insights, with workflows orchestrated in Airflow and scalable deployments via Docker and Kubernetes.

Created a platform that uses AI to manage online communities by moderating discussions, answering questions, and providing performance insights.

Built a multi-agent AI research assistant designed to support academic workflows, including literature search, paper drafting, and report generation. The system orchestrates autonomous agents to retrieve scientific knowledge from sources such as ArXiv and the web, index and search documents using Elasticsearch, manage research assets via object storage, and generate structured academic content using open-source LLMs.