New York, New York, United States
CAIO with 9 years architecting production AI systems: 4 years in DARPA ML research (autonomous systems, competency-aware models) + 5 years leading engineering teams building LLM-powered platforms at scale. Deep expertise in GenAI/LLM systems (RAG pipelines, agent-based frameworks, multimodal AI), production MLOps (1,250+ predictions/min, 500M+ records), and distributed systems. Expertise in finance, healthcare, defense, and blockchain domains.
Monstro is the operating system for governed financial intelligence, helping financial institutions deploy AI safely, explainably, and at scale. We’re building the infrastructure that enables more personalized and responsible financial guidance for millions of people.
- Devised agent driven ML system to predict industry codes for 120m+ businesses with 70% accuracy. - Shaped and led development of ELT pipelines for business data across 50 states. Shrank code by 80% (2k -> 400 lines), achieved 10x speed (sql -> dataflow), ingested 500M+ business records, 80% fewer data errors. - Constructed algorithmic framework to link domestic and foreign businesses, increasing accuracy from 50% to 80%. - Migrated core search infrastructure to a document database to slash search latency by 300%. Supported 2M+ searches to date, roughly 10k daily.
- Technical lead on a team of 8 senior engineers in implementing AI driven digital forensics tools for law enforcement to lower investigation time required to search seized laptops and phones. - Engineered SOC2 compliant agent-based framework to assist in criminal investigation. Formulated hybrid cloud and local architecture, robust error handling to ensure 99% uptime. - Transitioned to multimodal LLMs across 38 file types. Conducted systematic model evaluation for stability and back-testing to ensure reliability, increased accuracy by 35%. - Migrated cloud-based ML infra to local environments, adapting inference systems for GPU-less operations. Curbed system load by over 30% while maintaining accuracy requirements by pre-filtering files.
- Guided team of 8 engineers to implement anti-fraud tools. Integrated Quantexa into a Tier 1 Insurance company, cut false positive rates by 40%, rolled out database migrations to 38M users using distributed systems. - Blueprinted production RAG system for fraud investigation. Trimmed investigation by 35% while enabling analysts to access complex insights without advanced training. - Managed a full re-write of Quantexa Machine Learning Academy used by 2,000 engineers annually. Shaved training time by over 50% and trained all North American and Canadian new graduate hires on entity resolution best practices. - Headed development of internal developer tools to enhance entity resolution performance and KYC algorithms to lessen tuning time by over 60%.
- Deployed real-time cardiac monitoring ML models analyzing 10,000+ medical implants with <100ms latency, achieving 99.7% uptime for FDA-regulated devices through automated failover mechanisms. - Authored 40-page ML requirements document guiding 8 engineers, accelerating cloud infrastructure and model deployment timeline by 3 weeks through standardized best practices. - Designed Federated Learning hub-and-spoke platform to train models across 3 clinical applications in HIPAA-compliant manner, leveraging 300% more patient data from partner institutions without centralizing PHI.