Hands-on AI & engineering leadership
Own the difficult technology mandate.
I take on interim, operating-partner and, where the fit is right, permanent leadership across AI systems, platforms and teams. Here you can see the responsibilities I take on, the decisions behind my work, and the authority a mandate needs to succeed.
- years building production systems
- 18+
- years building production systems
- companies founded
- 6
- companies founded
- users on ML systems I led at Hike
- 100M+
- users on ML systems I led at Hike
- provisional patent filings
- 90+
- provisional patent filings
Dipankar Sarkar is a hands-on AI and engineering leader with 18+ years building production systems. He takes on accountable leadership mandates (hands-on CTO, Head of AI, VP Engineering, principal AI architect) and still ships code. He founded Neul Labs.
Leadership mandates
Different responsibilities, not different versions of a CV.
What separates these roles is authority: who decides, who answers for the outcome, and who sponsors the role. Pick the one closest to the gap you have.
All mandates and situations- 01 Hands-on CTO Own the technology direction and stay close enough to the implementation to interrogate it. Discuss
- 02 Head of AI One accountable owner across AI architecture, delivery, release decisions and the AI team. Discuss
- 03 Principal AI Architect Cross-team technical authority over AI standards and design, without line management. Discuss
- 04 VP Engineering Delivery cadence and engineering organisation for AI-heavy products, close to the architecture. Discuss
- 05 Interim technology leadership A critical function without an owner: stabilise it, make the decisions, hand it over. Discuss
- 06 Technical cofounder or operating partner Real operating ownership, with capital, equity, decision rights and commitment discussed openly. Discuss
Before you start a conversation
Latest Writings
Thoughts on systems, AI, and building teams.
Designing an On-Call Rotation That Works
Most on-call rotations burn out the engineers who are best at incident response, because the rotation punishes competence. Here is how to design one that doesn't.
Observability for AI Agents in Production
Standard APM tells you a request was slow. It doesn't tell you why an agent picked one tool over another. Here is what to log, trace, and alert on instead.
Trust Boundaries for Multi-Agent Systems
Single-agent guardrails like the Substrate Pattern don't compose automatically once agents delegate to other agents. Here is where multi-agent trust actually breaks.
Open Source
Tools, research, and infrastructure I build in the open.
Chain of Thought reasoning API with RL techniques
GDELT Project CLI — built for agents, optimized for automation
Your Guide to AI-Driven Business Transformation
Google Sheets CLI with local DuckDB caching and MCP server
Batteries included for iced (Rust GUI toolkit)
Your AI Agent's Gateway to HubSpot CRM
Built at & shipped with
Start a conversation
Have a mandate that needs an owner?
Share the problem, the authority on offer and the timeframe. Substantial interim and operating-partner mandates, and permanent roles where the fit is right. Compensation and confidential details can wait until we talk.
Contract AI engineering or a fixed-scope project?
That's dipankar.co.