New York, New York, United States
Building the system of record for onchain finance at Allium. Blockchains are powerful because they're permissionless, decentralized, neutral, and immutable. Those same properties make them unusable as financial infrastructure. They record what code ran, but not what it means. The raw data from an onchain transaction can represent a swap, a loan, or a liquidation depending on context. Understanding a single point-in-time balance can require replaying the chain's full history. The chain guarantees activity is recorded, but no one is accountable for how the data gets interpreted. That's manageable when crypto is self-contained. It breaks when institutions need to report positions, reconcile transactions, or close their books. I spent 8 years building distributed data systems in AI and NLP. In 2021, I started working with blockchain data and saw the same scaling problems I'd spent my career on: the biggest, most public dataset ever created, but no one had made it readable for institutional finance. Today, Allium translates raw blockchain events into standardized financial data across 140+ chains (SOC 1 & 2 certified). We power data infrastructure for 100+ companies like Visa, Coinbase, a16z, Phantom, and Metamask. We're building the data layer institutions trust not just for delivery, but for the definitions and methodologies behind every number. If your team needs onchain data you can analyze, build, and report on, let's talk: https://www.allium.so/contact See how we define and measure stablecoin activity with Visa: https://visaonchainanalytics.com/
Blockchain data cleaned, enriched and tailored to your needs: [Applications] Building DApps for Engineering teams [Analytics] Powering Analytics intelligence for growth and investment teams [Accounting] Data reconciliation for Finance teams We work with institutions & enterprises as their dedicated data infrastructure provider: VISA, Stripe, Phantom, Uniswap, DeFiLlama. Backed by Theory Ventures, Kleiner Perkins, Amplify Partners
Founded the Primer Command product and team, grew and led 20+ engineers across all applications on the platform. We ingested and processed the world's twitter data, news data, telegram data in near realtime to serve mission critical intelligence needs for various institutions. * Built Primer's core algorithms - GPU Realtime Clustering at Scale (presented at Nvidia GTC 2020 on using sparse matrix multiplication on CUDA to optimize pairwise comparisons for Louvain clustering) * Built Semantic Search in 2020 (now know as RAG) * Active Learning based Information Retrieval system 2018
Lectured CS102 Data Mining + Machine Learning with Lisa Wang. This experience taught me how to teach.
Built a traffic management system for unmanned aircraft based on Reinforcement Learning, and carried out physical flight tests demonstrating the algorithm's effectiveness with Prof Mykel Kochenderfer
Head TA for CS238: Decision Making Under Uncertainty by Prof. Mykel Kochenderfer TA CS 228: Probabilistic Graphical Models with Prof. Ermon
Built predictive machine learning models to grow Azure
Built SparkAid - Auto tune performance of Apache Spark apps. Presented @ ApacheCon '15