Building things with
AI that actually work.

Built an agentic AI system that automates job discovery, semantic search, ranking, and summarization. Designed a retrieval-augmented pipeline that ingests job postings, generates embeddings, and recommends relevant opportunities. Reduced job research time by 60%.

Developed an AI-powered support assistant for troubleshooting appliance issues, retrieving repair documentation, and recommending replacement parts. Indexed 1,100+ parts and guides with a multi-tool RAG agent, built end-to-end in 48 hours.

Designed an AI-driven applicant intake system that analyzes backgrounds, classifies skills and interests, and recommends suitable opportunities through automated routing workflows. Evaluated using precision, recall, and macro F1.

Built a cloud-based text-to-SQL analytics platform on Snowflake using U.S. Census data, powered by Cortex Agent for natural language querying over demographic datasets. Developed SQL transformation pipelines, dimensional data models, and Python ETL workflows to ingest, validate, and load millions of records, enabling conversational exploration of population, income, education, and geographic trends.

Designed a scalable pipeline for processing large-scale satellite datasets. Automated ingestion, transformation, validation, feature engineering, and model training with Spark on AWS. Reduced processing time from hours to minutes.

Built a lightweight SQL engine capable of parsing and executing SQL-style queries on structured datasets, implementing filtering, sorting, aggregation, joins, and relational algebra fundamentals.

Built an NLP platform for legal document classification and summarization using transformer-based models and ML classifiers. Interactive Streamlit app achieving 96.9% classification accuracy.

Developed an end-to-end computer vision pipeline for automatic license plate recognition, combining deep learning classification, OCR text extraction, and synthetic data augmentation.

Built statistical forecasting models to analyze volatility and correlations across financial assets using GARCH and DCC-GARCH econometric techniques with rolling forecast evaluation.