AI products after the demo
What actually decides whether a model-backed product becomes useful: evals, cost, latency, trust, workflow, and taste.
Writing & Speaking
AI products in practice.
I write and speak about what happens after an AI demo works: evaluation, product quality, developer tools, cost, reliability, and the leadership decisions behind useful systems.
Talks
Talks for product and engineering teams.
These fit engineering, product, founder, and operator audiences that want concrete AI product lessons rather than a broad trend talk.
Evaluation as product work
How teams can use evaluation to make product and engineering decisions, not just produce a score.
AI and engineering leadership
How AI changes review, ownership, technical direction, and the leadership work around software teams.
Code review in the AI era
What code review has to become when teams can create more software than they can confidently understand.
Selected writing
Selected writing.
A few essays and notes on building AI systems, learning from prototypes, and turning technical work into product judgment.
From Vibecoded Project to OpenAI: Why the OpenClaw Move Matters
What the OpenClaw story suggests about long-running agents, and why autonomy only becomes useful when it is shaped by workflows, constraints, and feedback loops.
6 min read
AI Consulting: Lessons from a Year of Thinking, Talking, and Building
Lessons from a year of AI consulting conversations: why trust, framing, and business understanding matter as much as the technical solution.
10 min read
Speaker kit
Bios for organizers.
Short bio
Gaurav Pooniwala is an AI engineer and founding engineer at Cubic. He has worked on generative video, AI code review, agentic workflows, retrieval systems, and production ML, with a focus on making AI products useful beyond the demo.
Longer bio
Gaurav Pooniwala is an AI engineer and founding engineer at Cubic, where he leads AI development for code review. His career spans deep learning research at Samsung R&D, generative video systems at Rephrase AI before its acquisition by Adobe, agentic workflow products at startups, and current work on AI systems for software teams. He writes and speaks about evaluation, product quality, developer tools, technical leadership, and the work required to make AI useful in real organizations.
Start a conversation
For talks, panels, podcasts, workshops, essays, or collaborations, send a note with the audience and goal.