Suhaas Gaddala

I build AI agents at Blaze that move real money on stablecoin rails.

I was one of the main developers of Blaze's AI CFO, an agent that runs financial operations for businesses, from checking balances to sending cross border transfers.

I build the production TypeScript and Express systems that let agents discover, pay for, and call x402 priced model and data APIs with USDC settlement. That work covers provider routing, payment clients, safety caps, and cost controls, with integrations across Anthropic, Coinbase CDP, DexScreener, and Bazaar, all backed by 125+ tests.

I also designed and shipped the usage accounting layer that traces cost, margin, and ownership across every provider call an agent makes.

I architect machine learning systems for autism-related behavioral phenotyping in controlled social interaction experiments. The work focuses on converting raw behavioral data into measurable interaction signatures, delivering 88%+ accuracy across 150+ experimental trials, and building a high-throughput behavioral intelligence pipeline that reduced preprocessing time by 76%+ through automated annotation, feature extraction, and downstream analysis.

I built cloud-native backend services and infrastructure tooling across distributed systems at Dell, working in Go and Python while integrating Kafka and MQTT to improve event processing throughput by 40%. I containerized services with Docker and Kubernetes, reduced deployment cycles from hours to minutes, and implemented Vault, Prometheus, Jaeger, and Fluentd across 10+ distributed services to improve secrets management, observability, trace depth, triage speed, and production reliability visibility.

I built and improved digital infrastructure for accessibility-focused fundraising and outreach across the United States and India. I helped scale monthly traffic from roughly 800 to 12,500+ visitors, redesigned donor onboarding flows that increased conversion rates by 3.4x, built targeted outreach systems that expanded program visibility to 40,000+ families, and streamlined recurring fundraising operations so campaign execution moved about 55% faster.

I worked across autonomy, perception, and data infrastructure for Nova, UT Dallas's applied autonomous driving project. I engineered a multimodal pipeline processing 1.2M+ LiDAR, vision, and GPS frames per week, reduced model training time by 47%, and built a LiDAR-camera route-risk model that improved hazard precision by 38% while reducing false positives by 52%.