Keep pulling the thread on Varun.
GigaML raised a $4 million seed round after gaining traction from its open-source models.
GigaML, as a team of 8 people, successfully won a contract with DoorDash by competing against a well-funded company with 400 employees.
GigaML is developing a new product, an "AI forward deployed engineer," designed to automate the configuration and policy changes required for enterprise AI deployments.
Varun estimates that GigaML would need 6 to 7 times more engineers if the company did not use AI coding agents.
Varun claims that Anthropic and OpenAI do not pay commissions to their salespeople.
GigaML builds AI agents for customer support and counts DoorDash, a top-three global telecom provider, and a major crypto exchange among its clients.
GigaML aims to achieve customer support deflection rates of 90% to 95% for its top customers using its AI agent technology.
GigaML's initial idea when applying to Y Combinator was to build an edtech product using large language models.
Y Combinator partner Harj Taggar advised the GigaML founders that their edtech idea would not work and encouraged them to pivot to a different concept.
GigaML gained initial traction by open-sourcing several models that achieved top rankings on Hugging Face benchmarks.
GigaML's co-founder Varun believes the market for LLM fine-tuning services is poor because its primary use cases are limited to cost reduction and speed improvement.
GigaML discovered its current focus on customer support after observing that customer support and coding were the only two rapidly growing use cases among its fine-tuning customers.