"By far, the greatest danger of Artificial Intelligence is that people conclude too early that they understand it."
My work focuses on designing and deploying large-scale machine learning and GenAI systems, with an emphasis on LLM-powered agents, retrieval systems, and production AI infrastructure. I am broadly interested in building reliable, scalable, and impactful AI systems that operate in real-world environments, spanning areas such as generative AI, multimodal systems, and intelligent automation. Below are some of my selected projects. If anything is of interest, feel free to contact me!
Selected work
01
Designed and deployed a production-grade GenAI agent platform integrating retrieval, tool usage, and code execution. Built scalable pipelines leveraging BigQuery and LLM orchestration frameworks to enable complex multi-step reasoning workflows. Improved response quality, grounding, and system reliability for enterprise use cases.
02
Developed a multimodal AI agent for hazard detection from store ceiling images, combining visual understanding with reasoning and task automation. Implemented localization and decision-making pipelines that automatically assign tasks to store associates, enabling real-world operational impact. Focused on robustness, interpretability, and safety in production environments.
03
Built and deployed an AI-powered semantic search system for call center agents using LLMs and retrieval pipelines. Reduced response time by 60% and significantly improved knowledge accessibility and agent efficiency. Designed end-to-end system including data ingestion, indexing, retrieval, and response generation.
04
Developed a recommendation system to assist call center agents with next-best actions, reducing average call duration by 20%. Leveraged historical interaction data and machine learning models to optimize decision-making in real time. Integrated into production workflows with measurable business impact.
05
Led the design and implementation of a large-scale data pipeline processing global call center interactions. Built NLP-based systems for sentiment analysis, issue classification, and trend detection to support operational insights. Enabled real-time monitoring and improved decision-making for leadership teams.
06
Developed a reliability scoring model for electric vehicle charging stations and optimized home charging costs using intelligent power system modeling. Combined statistical modeling and optimization techniques to improve system efficiency and user experience.