Abhijit “Roy” Barman
I am a hands-on CTO and full-stack architect based in San Francisco. I build alongside the team, not only above it.
- Currently fractional CTO at an AI-native mental-health startup — took the platform from zero to millions of DAU with SOC2 / HIPAA / NIST AI RMF compliance.
- Previously CTO & VP Engineering at BCT Group: scaled engineering from 150 → 400+ and ARR from $20M → $120M+ in three years.
- Cofounded a rideshare marketplace; led carrier-grade location systems used by AT&T, Verizon, T-Mobile, Vodafone, and others; owned a $50M+ ARR fintech P&L serving the largest U.S. banks.
- 8+ years shipping production AI platforms for healthcare, fintech, and transportation.
- Built and led global teams across the US, India, Taiwan, UAE, and Slovakia.
- U.S. Citizen. Based in San Francisco.
If you think I may be helpful in a specific way, get in touch by email or LinkedIn.
Selected work
CynosureConnect — Fractional CTO
Built the hybrid multi-agent platform from the ground up on GCP Vertex AI, Ray on GKE, and Temporal. Connected partner systems via MCP servers. Designed the core RAG pipeline (hybrid retrieval + cross-encoder reranking) and full LLM evals / observability. Shipped a human-in-the-loop mobile + voice product that is SOC2 / HIPAA / NIST AI RMF compliant.
Outcome: zero → millions of DAU with thousands of concurrent users.
BCT Group — CTO & VP of Engineering
Inherited a 150-person team and built an AI-native engineering organization across the US, Taiwan, UAE, and India. Formed an Architecture Council, drove SOC2 / HIPAA / GDPR / CCPA as a sales enabler, and rolled out AI pair-programming (Copilot, Cursor, multi-model agents) that tripled velocity and cleared years of tech debt.
Outcome: 150 → 400+ engineers · $20M → $120M+ ARR · 100+ releases/year · products used by Tesla, NFL teams, government & telco accounts.
Fijo Technologies — CTO & Cofounder
Took the company from idea to production native iOS and Android apps on a Firebase serverless stack, with Checkr / Veriff / Coinbase / PayPal for trust infrastructure and a TensorFlow pipeline on Vertex AI.
Polaris Wireless — Senior Director, Mobile Products
Led global engineering for a sensor-fusion indoor/outdoor location engine that requires no special network hardware or device modification — deployed by AT&T, Verizon, T-Mobile / Deutsche Telekom, Swisscom, Vodafone, and Orange. Improved a key customer’s NPS from −90 to +90 and helped secure a $70M contract. Stood up a delivery center in Slovakia that cut global ops cost 40%.
CoreLogic Dorado — Executive Product Director
Owned P&L for a $50M+ ARR platform serving JPMorgan Chase, Bank of America, Wells Fargo, and Citibank. Re-architected a multi-petabyte Oracle system for 10,000+ concurrent users with sub-second SLAs. Invented a parallel processing algorithm that saved JPMorgan Chase $300M+ per year.
Outcome: $150B+ loans processed per year · NPS −50 → +90 · ~20% annual revenue growth.
NaviSite (Accenture) — Director, Cloud Architecture
Led engineering across 16 data centers and delivered $100M+ in cloud transformation projects. Reduced Granite Construction’s TCO by 70% ($16M migration, 1,000+ servers → ~100 VPS) and improved BAE Systems’ ROI 3×.
Oracle — Practice Manager, Emerging Technologies
Built a 50+ person architecture practice from zero, serving 200+ enterprise customers (Cisco, Walmart, Salesforce, Genentech, HP, Intuit, and many others) at ~99% CSAT. Raised utilization from 70% → 90%+ and profitability from 20% → 40%+.
How I work
- Build with the team, not above it. I still write production code and open PRs.
- 0 → 1, then 1 → 100. Greenfield hard problems and the messy middle of scaling them without losing the P&L.
- Hybrid agents + deterministic code. LLMs are powerful and expensive. I design systems that use both so cost and reliability stay under control.
- Compliance as a feature. SOC2, HIPAA, GDPR, NIST AI RMF — treated as product capabilities, not paperwork after the first enterprise deal.
- Global teams, one standard. Same quality bar across US, India, Taiwan, UAE, and Slovakia.
What I reach for
AI / agents: Temporal, LangGraph, MCP servers, Ray, FastAPI, Cursor, Claude, Copilot, Langfuse, hybrid RAG + reranking
Models: Anthropic Claude, OpenAI GPT, Google Gemini, AWS Nova, Mistral
Languages & data: Python, Go, TypeScript, Java, Rust · Postgres + pgvector, TimescaleDB, Oracle, Redis, Snowflake, Databricks
Cloud & ops: GCP (Vertex, GKE), AWS, Azure · Kubernetes, Terraform, Prometheus / Grafana / Loki, GitHub Actions
Beyond the day job
- MIT Graduate Program — Robotics & Quantum Computing (continuing)
- UC Berkeley — AI / ML Full Stack Deep Learning
- AWS Startup CTO Fellowship
- MBA, Finance & General Management — UC Davis
- B.E. Computer Science, Electronics & Telecom — Jadavpur University
- Advisor: Par3 Group (CEO), OMG / Digital Twin Consortium (CSO & CTO)
- Mentor, Global CTO Forum · ACM SIGAI · IEEE Quantum Technical Committee
- Patents: Unified Risk Signal Architecture (mental health); Trusted rideshare models
- Volunteer tech & coding coach — Salvation Army, San Francisco
Get in touch
Especially interested in AI platforms that have to work in the real world — regulated domains, high scale, or both.
roy.barman@gmail.com
linkedin.com/in/abhijitroybarman
+1 (415) 786-6637