Small team. Senior bench. Humans on every decision that matters.
We do not staff junior engineers on production AI systems. Every engagement is led by a principal who has shipped this kind of system before, and who stays involved through Operate.
Deep RAG and multi-agent practice. Built Newsweek’s production article-to-video pipeline (95% success at $271 / 1K videos), N1Healthcare’s end-to-end medical-report AI with 20+ specialised agents over 100K+ LOINC codes, and CGIAR’s GPU-accelerated PDF extraction across 9K+ research records. INSPIRE scholarship recipient.
Expertise
RAG · Vector searchMulti-agent systemsMulti-LLM orchestrationDocument OCR pipelinesClinical / medical AIAzure AI Foundry
PE-backed data systems, business automation, executive analytics
Six years architecting intelligent systems across PE-backed portfolio operations. Partnered with Incline Equity Partners and Riverside Company, plus enterprise analytics for Google, Pfizer, Novartis, Lilly, and Delta. Typical impact: 15% operational efficiency, 20% marketing ROI, 10–40% manual-task reduction.
Autonomous agents, multi-model routing, production AI systems
Six years building production AI. 600+ automations shipped across 35 projects in 12 industries. Creator of Kronus, an open-source agent OS with 21 specialised agents, swarm mode, and a multi-tenant Hub. Founder of phavella. Systems have lifted LLM accuracy 15–25% on benchmarks and held 99.9% uptime in production.
Hybrid product–engineering lead at the intersection of GenAI and financial risk. Founder of Swiftly, a multi-agent platform orchestrating LLM + deterministic components for extraction, reasoning, and generation. Previously led $150M portfolio risk strategy at Drip Capital, designed a $1B / year unified transaction policy, and built neural-network fraud models at Citi delivering $1.4M–$15.6M annual savings.