Resume

Matter42 · Scientific ML / Materials metrology / Python

• Building a simulation-to-ML pipeline that generates synthetic training data for predicting vacancy defect densities in 2D materials, starting with WSe₂ monolayers
• Benchmarking machine-learned interatomic potentials (ORB v3, MACE) against DFT phonon references, using κSRME to select models with physically reliable force constants
• Computing phonon spectra with Phonopy as Raman proxies and simulating HRTEM/STEM images with abTEM to train defect-segmentation models
• Developing models that infer defect density from Raman features, so labs can skip slow, expensive TEM for routine defect characterization

April 2025

• My first exit: Nousot, where I started as an ML engineering intern and later led the creation of its self-service forecasting platform, merged into Lovelytics
• The combined firm formed a 500-person, Databricks-focused data and AI consultancy operating under the Lovelytics name

Ouro · GenAI / Full-stack / Agents / Materials science

• Founded and built an AI-first technology platform, architecting full-stack implementation including frontend, backend, design, and integrated AI capabilities
• Built ouro-agents, an open-source framework for long-lived autonomous agents, and a team of scientist agents that run hypothesis-driven materials discovery programs
• Partnering with the Snyder Group at Northwestern University on permanent magnet research, pairing Ouro's computational discovery pipeline with the group's materials expertise
• Deployed 15+ materials science services on the platform, including a DFT property API (ABACUS + TB2J), thermoelectric ZT prediction, and generative crystal models, powering an open search for rare-earth-free permanent magnets
• Developed and deployed an advanced AI agent system with custom tooling integration, enabling real-time platform assistance through deep integration with documentation, user data, and platform capabilities
• Pioneered an intelligent support architecture combining RAG-based knowledge retrieval with custom API tooling, allowing users to instantly access and act on platform-wide information
• Built and deployed enterprise-grade LLM streaming infrastructure, enabling real-time AI responses for developers and end-users while maintaining high performance and reliability at scale

Meta

• Architected and deployed an end-to-end ML forecasting pipeline from scratch, achieving 20% improved accuracy in predicting resource utilization across compute, storage, and specialized processing resources, enabling millions in cost savings
• Led development of deep learning forecasting models to optimize resource allocation across Meta's global infrastructure, preventing processing bottlenecks and enabling proactive capacity planning
• Designed and implemented automated data cleaning and preparation workflows to process multi-dimensional resource usage data for diverse resource types including compute, storage, and video processing

Nousot

• Fine-tuned LLMs for structured data extraction, saving analysts 100s of hours and cutting costs by 90%
• Worked with Facebook building an internal tool that saves data scientists across the FAST, FAM, Growth, and Gaming teams weeks of work
• Led the creation of the company's latest product offering — a self-service forecasting and data platform with a frontend, backend, and forecasting API

Nousot

• Managed day-to-day aspects of building custom software for a multi-billion dollar company in an Agile environment
• Implemented scripts to automate building complex GeoJSON shape files based on the client's distributor boundaries
• Contributed to a web scraping project that builds datasets ready for the application of machine learning algorithms

Northwestern University

• Weinberg College of Arts & Sciences, Undergraduate
• Balanced a demanding academic curriculum with an equally demanding 20+ hour/week commitment to athletics
• Cultivated a winning culture within the team, leading with 16 points and the best season finish in over 5 years