Why Two IIT Engineers Turned Down $550K Jobs To Build A Startup

GigaML builds AI agents for customer support and works with companies including DoorDash. Varun says traditional support systems deflect about 10 to 15% of requests, compared with 60 to 70% for AI, with a goal of reaching 90 to 95% for top customers.

Varun grew up in a small town in Andhra Pradesh, studied electrical engineering at IIT Kharagpur, and began LLM research during college. He and his co-founder initially pursued an EdTech idea, but YC partner Harj encouraged them to choose another direction. They later built fine-tuning tools, released open-source models, and raised a $4 million seed round.

After about a year, the founders found that customer support and coding were the strongest use cases among their fine-tuning customers. They focused on support, with Zepto as their first customer. Varun says their eight-person team later won DoorDash’s business after a three-month pilot, which helped build trust with other large companies.

Varun says enterprise AI agents depend on iterating on policies to improve measurable outcomes such as resolution rate and customer satisfaction. He advises founders to test whether customers will pay before building extensively, and says GigaML now seeks customer commitments for new products. He also says the company is developing an AI forward-deployed engineer to help customers make policy and dashboard changes.

GigaML encourages employees to automate work, including using OpenClaw to schedule meetings and Claude Code to build tools for analyzing sales call transcripts. Varun estimates the company would need six to seven times as many engineers without coding agents, while stressing that understanding code remains important. Its interviews ask candidates to write code with AI, then modify it without the tool.