I'm Anurag — a Forward Deployed Engineer and the founding engineer at Debales AI. For the last two and a half years I've built AI products end to end: the agents that talk to customers, the integrations that put AI inside the tools merchants already work in, and the systems that keep all of it running under real load.
That's meant wearing whatever hat the work needs. Some weeks I'm deep in backend architecture, designing systems that process high volumes of events without falling over. Other weeks I'm operating cloud infrastructure, fixing what observability surfaces, and making sure the platform holds up. Lately, more of my time has shifted to working directly with customers — sitting in their calls, understanding what's actually broken in their day, and turning that into product.
I came up leading a small engineering team along the way, which taught me the thing I keep coming back to: the best technical work isn't the cleverest code, it's the work that unblocks other people and survives contact with reality. The wins I'm proudest of look unglamorous on paper — a backend rewrite that cut infrastructure costs, a monorepo migration that dropped deployment issues, an integration that lifted onboarding. None are interesting to talk about at a party. All of them moved the business.
What excites me now is how much of AI's real value lives in the integration layer — the edge cases, the customer-specific quirks, the loop between what someone says they need and what they actually need. That's the work I want to keep doing.