Mehak
Aggarwal

The algorithm that Mehak Aggarwal built at Harvard could find fractures in bone. Hairline cracks invisible to the eye, buried in CT scan data, surfaced by pattern recognition trained on thousands of cases. The paper got 406 citations. The technology got patented. And then she walked away from all of it to build something that finds cracks in sales pipelines instead.

That pivot - from medical imaging to sales AI - makes more sense when you understand how Mehak thinks. She's not someone who chases problems that are impressive. She chases problems that are specific. At Harvard's Athinoula A. Martinos Center for Biomedical Imaging, the specific problem was: how do you make diagnostic AI that doctors can actually trust? At Sybill, the specific problem is: how do you give salespeople a second brain that works in real time?

The company started in 2020. Four roommates. Ramen. Late-night coding sessions in Mountain View. Gorish Aggarwal - her brother, now CEO - had noticed something strange while teaching at Stanford during COVID. On Zoom, you can't read a room. The raised eyebrow. The slightly-too-long pause before someone says "that sounds interesting." The facial tells that tell you whether a deal is real. Gorish thought the problem was his. Mehak recognized it was everyone's.

What she built is not a transcription tool. Transcription is the easy part. Sybill analyzes the full arc of a deal - every call, every email, every moment where a buyer's tone shifts or a competitor gets mentioned - and synthesizes it into something a rep can use immediately. Post-call summaries written in seconds. CRM fields updated automatically. Follow-up emails drafted before the rep has even closed their laptop. The AI doesn't just record what happened; it tells you what it means.

The research background shows. Sybill's approach to multimodal AI - pulling signal from speech, text, and non-verbal cues simultaneously - maps directly onto the methods Mehak developed in healthcare. In medical imaging, you don't get to pick one modality. You use MRI, CT, and clinical notes together, because the fracture often only appears when you triangulate. In sales, the deal risk often only appears when you read the call transcript alongside the CRM history alongside the email response time. Same logic. Different domain.

The numbers from the first 18 months were not subtle. Sybill went from $100,000 to $1 million in annual recurring revenue in 9 months. Then it grew 15x from there. More than 60% of new customers came through referrals - the product sells itself when a sales rep realizes they're saving 5 hours a week and their CRM is actually accurate for the first time in years.

The Series A came in July 2024. $11 million, led by Greycroft, with Neotribe Ventures, Powerhouse Ventures, and Uncorrelated Ventures in the round. Total funding hit $14.6 million. The team doubled. Sybill expanded to 30+ countries.

And then, in 2025, Mehak added something else to her schedule: a fully funded MBA at Stanford's Graduate School of Business, as a Knight-Hennessy Scholar. Knight-Hennessy is Stanford's flagship graduate scholarship program - roughly 100 scholars per year, selected for leadership, purpose, and impact. Mehak wrote about accepting it on LinkedIn, and the post got 763 reactions. Not because she was celebrating a credential. Because of what she said about the journey that got her there.

She grew up in Chandigarh, India. Did a dual-degree at IIT Delhi - B.Tech and M.Tech in Mathematics and Computing, one of the most competitive programs in the country. Graduated in 2021. Somewhere in between, she was doing research at the Singapore University of Technology and Design on low-cost cyberattack defense systems, and at Hebrew University of Jerusalem on AI for fracture detection with Hadassah Israeli Hospital. These aren't the resumes people write to sound impressive. They're the resumes of someone who genuinely likes hard problems and moves toward them.

As CPO and Head of AI at Sybill, Mehak owns the product roadmap and the core AI architecture. That's an unusual combination - product thinking and deep technical execution in the same person. It means Sybill's features don't get built for the demo. They get built to work reliably, at scale, in the hands of sales reps who will abandon the tool the moment it gets something wrong. "If the AI output is not accurate, people lose trust in the system very, very quickly," her co-founder Gorish told TechCrunch. Mehak is the one making sure that doesn't happen.

She spoke at Princeton's Language and Intelligence Workshop in August 2024, specifically on reliable AI agents - the hard problem of getting AI systems to do what they say they'll do, consistently, without hallucinating or drifting. It's the same problem she worked on in medical AI, where a confident-sounding wrong answer has consequences. Sales AI is lower stakes than radiology. But the design principle is identical: the system has to earn trust, over time, by being right.

Stanford isn't a finish line for her. She was clear about that. She's going to keep building Sybill - with, as she put it, "the same scrappy energy that started it all." The MBA is a reflection point, not a landing pad. The dots, she wrote, connect later. She's still mid-stride.