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

Predictions and Expectations:

• Varun expects deflection rates for top customers to reach 90% to 95% with their AI agents.
• The speaker believes that the biggest bottleneck in enterprise AI deployment will be the "forward deployed engineer" role.
• The company is confident they will "take over" the forward deployed engineering function soon.
• Varun believes product is the most important factor for successful AI companies, outweighing sales teams or sales commissions.
• The speaker states that "if it's an important enough problem, people would pay" money for it, or time in the case of social media.
• Varun predicts that without coding agents, the company would need six to seven times more engineers than it currently does.

Timelines and Milestones:

• The company plans to launch their "AI forward deployed engineering" solution soon.
• Varun previously told his parents that even if the startup fails in "one or two years," he could return to a job.
• The team piloted with DoorDash for three months before securing the contract.
• The company took "about a year" to realize that fine-tuning was a bad market due to sales process complexities.

Technology and Product Direction:

• GigaML is pivoting its mission to become a "generic automation builder" able to automate anything on top of their platform.
• The company intends to build an "AI forward deployed engineering" solution that automatically joins meetings, takes notes, and implements policy changes.
• The product roadmap involves iterating on markdown files to improve business KPIs like resolution rates and CSAT.
• The speaker plans to automate all internal work, including sales analysis using transcripts and meeting scheduling via Cloud Runs.
• The company aims to move from a 30% to 40% resolution rate up to 90% by iteratively improving their markdown-based policies.

Market and Industry Outlook:

• Varun believes innovation in the Gen AI field is driven almost entirely by the Bay Area, making San Francisco the necessary location for research-based work.
• The speaker expects the market to shift toward enterprises buying software from small startups rather than assuming only large companies are trustworthy vendors.
• The company predicts that "sales" is no longer the primary success driver for AI companies, as successful firms like Anthropic do not use traditional sales incentives.
• Varun observes that customer support and coding are the only two high-growth use cases for their fine-tuning technology.

Company Plans:

• The team is intentionally moving toward building generic automation capabilities rather than focusing solely on customer support.
• The company plans to hire "extraordinary ability" and "spikiness" in candidates, specifically looking for individuals in the top 0.1% of their fields.
• Future hiring processes will continue to involve testing candidates' ability to change code without AI access to ensure deep code understanding.
• The company plans to target Fortune 500 companies for internal support and compliance automation pilots.
• The founders are denying acquisition offers to continue scaling and reaching their full potential.

Financial Guidance:

• Varun mentions raising a $4 million seed round prior to their pivot.
• The company previously offered a quant job to the speaker at $550,000, which was turned down to pursue the startup.
• The speaker earned approximately $50,000 from Kaggle competitions to fund early high-frequency trading jobs.

Risks and Caveats:

• The speaker acknowledges the risk of burning boats by rejecting high-paying jobs without a secured business model.
• Varun admits they initially worked on "stupid ideas" that generated no revenue for a long time.
• The speaker notes that fine-tuning became a bad market because selling to secure sectors like insurance and healthcare is an engineering sales process, not just an engineering task.
• The company faced a risk of rejection during YC where both B1 and B2 meetings were initially rejected.

Confidence and Disagreement:

• Varun states he is "very confident" that the company is moving in a great direction regarding AI adoption solutions.
• The speaker expresses strong confidence in the "forward deployed engineering" solution they are building to solve enterprise bottlenecks.
• Varun admits he was "so wrong" about the importance of sales, contrasting his initial belief with the reality of AI product-led growth.
• The speaker claims that for AI companies, product value is the primary differentiator, noting that major competitors like Anthropic do not pay sales commissions.