Selected Works

Owning architecture and hands-on build of a GenAI suite spanning three marketing domains (category, media, brand) on Azure Databricks. Designed the cross-product AI integration layer (model routing, chat runtime, prompt and tool registries) that keeps agent behavior consistent as the suite scales across products. Engineered the agent harness with evaluation loops, guardrails, and observability, cut workflow bottlenecks across the suite, and drove back-end / front-end migration plans into the product roadmap together with PMs and stream architects as part of a wider AI transformation program.

Built an AI-agent ecosystem from zero on Azure Databricks to surface drivers of postpaid churn: multi-agent orchestration, intent routing, and a talk-to-your-data layer over the operator's customer and billing data. Applied loop and graph engineering (LangGraph-style stateful graphs with planner, executor, and critic nodes) so agents can iterate on hypotheses instead of answering in one shot. Implemented root-cause-analysis tooling the agents call to pinpoint churn factors and recommend countermeasures, integrated the agents into enterprise workflows, and tuned the system for accuracy, scalability, and reliability with eval-driven development.

Developed an internal enterprise chat system with secure access to private and public company documents. Leveraged RAG (Retrieval Augmented Generation) to provide accurate, context-aware answers based on proprietary data while maintaining strict access controls.

Built a custom VS Code extension powered by an LLM fine-tuned on the company's specific codebase. Significantly improved developer productivity by offering context-aware code suggestions, refactoring, and documentation generation tailored to internal standards.

Architected a 'Talk-to-your-data' system featuring a multi-agent orchestration layer. Users interact via chat to perform complex data analytics and causal inference tasks, with specialized agents handling SQL generation, statistical analysis, and visualization.

Led the development of an innovative feature integrating detected lanes and road markers into a comprehensive environmental map. Collaborated with the Context-Aware Scene Prediction Network to enrich navigational systems with precise, real-time scene understanding and prediction capabilities.

Led the creation of a high beam assistant feature using machine learning to identify the light from oncoming traffic, enabling automatic adjustment from high to low beam. Enhanced night driving safety and comfort, showcasing a practical application of AI in improving automotive lighting systems.

Led a team of 10 developers in designing a ML algorithm to enhance road safety by reducing false positive detections (shadows, projections, road markings). Significantly improved detection accuracy for autonomous driving technologies.

Led the development of an advanced real-time system for detecting and tracking road markings, leveraging machine learning for enhanced accuracy and safety in autonomous navigation technologies.

Directed technical development of a platform facilitating contractor-client matchmaking in the home-care sector. Incorporated dynamic pricing and workload forecasting to optimize service delivery.

Developed a novel software solution for dynamic road speed regulation based on real-time traffic management data, leveraging AI for public safety and efficiency.

Pioneered a smart factory management system integrating software-enhanced smart glasses to guide plant workers in selecting correct parts, tools, and assembly techniques. Intersection of AI and industrial IoT.