Work

Building AI from
prototype to production.

Selected projects spanning LLM agents, reward modeling, anomaly detection, and ML systems at scale. Each one shipped.

Microsoft · Fabric · 2024 LLM Agent

Operations Agent

Designed and built an LLM-powered agentic system for operational intelligence within Microsoft Fabric's Real-Time Intelligence platform. The agent autonomously investigates operational anomalies, identifies root causes, and recommends actions — all grounded in real telemetry data.

Technical depth

Built custom reward signals for agent evaluation, developed an auto-rater evaluation framework for measuring agent quality at scale, and designed the agentic architecture from retrieval through action generation.

Stack

Python · LLMs · Reward Modeling · Evaluation Framework · Azure

Microsoft · Fabric · 2023–24 ML System

Anomaly Detector

Built an AI-powered anomaly detection product for real-time streaming data within Fabric. Handles enterprise-grade telemetry with low-latency inference, interpretable outputs, and configurable sensitivity — designed to be both powerful and usable by non-ML practitioners.

Technical depth

Developed the detection algorithms, built the serving infrastructure for real-time inference, and designed the interpretability layer that translates model outputs into actionable insights.

Stack

Python · Statistical ML · Streaming · Real-Time Inference · Azure

Microsoft · Fabric · 2024 AI Product

Operational Discovery

Created an automated system that discovers operational patterns across time-series data, combining statistical methods with LLM-generated insights. Turns raw telemetry into structured, actionable intelligence without requiring manual exploration.

Technical depth

Hybrid approach combining classical time-series analysis with LLM-powered summarization and insight generation. Designed for scale and interpretability.

Stack

Python · Time Series · LLMs · Statistical Analysis · Azure

Previous experience

Before Microsoft, I held roles at Google, Pinterest, Netflix, and Meta spanning recommendation systems, experimentation platforms, search ranking, and content understanding. Each role deepened a different dimension of building ML systems at scale.