17 Young Founders | 8 Startups | 20 Lakh Grants | Third WTFund Cohort
Executive Summary
Eight WTFund Cohort 3 startups pitch Nikhil Kamath at his Goa home: Prava Payments (payment rails for AI agents), OncoALERT (Rs 150 needleless oral-cancer screening built for villages with no electricity), an AI operating system for diagnostic labs (digitized pathology slides plus biomarker-driven drug recommendations), antimattr (an agentic 'Android for earbuds'), O3 Security (end-to-end cybersecurity with context, already billing ~$80K in five months), BloomRehab (home rehab robots for stroke patients), Placestation (desktop pick-and-place machines that compress hardware iteration from two months to a day) and Aeyi (turning CCTVs into AI security guards). Between pitches, Nikhil gives the cohort his real playbook: learn storytelling because 'if you can't appeal to me in that one line, tomorrow, when you sell it, that's the same line you're gonna use'; learn negotiation because 'people don't often disagree with ideas, they disagree with people'; don't let family or society timetable your marriage because 'hypocrisy and change are synonymous'; and his own capital is going into energy transition, because AI or no AI, compute needs power.
Key Takeaways
AI agents can discover but can't pay - that's a business
Prava Payments' founders saw AI assistants halt at any action requiring money. Their PSP-agnostic stack lets an AI application onboard your card once and pay any merchant (Stripe, Adyen, Razorpay) without leaving the chat - turning a seven-step detour into three steps. They're targeting the US and Southeast Asia first, with UPI plans for India later.
Design for the village with no electricity, lab or staff
Dr. Jayanti's OncoALERT came from a rural Assam posting where oral-cancer screening is naked-eye visual inspection with 50-60% accuracy - 'as good as a flip of a coin' - against a policy that says everyone over 30 who uses tobacco should be screened (a 70-80 crore population). Her nanotech biosensor costs about Rs 150, is needleless, and was designed by asking: what if nothing works - no power, no lab, no trained people? Distribution runs B2B through hospitals and insurance.
Software distributes, kits don't
When Nikhil forced the room to bet $100 on one of two cancer startups, the pathology-AI team won on distribution logic: shipping physical test kits to medical stores is brutally hard, but digitizing existing microscope feeds and selling software to diagnostic labs scales anywhere. Their OS digitizes glass slides, runs AI algorithms in real time, and maps biomarkers to targeted drug therapies for cancer patients.
Voice is becoming the primary interface
antimattr's bet: earbuds and wearables become the default computing surface as voice replaces screens - an 'Android for earbuds' with integrations like Swiggy and Uber built in. The founders themselves have stopped typing: dictation tools like Wispr Flow raise words-per-minute, and even coding is increasingly done by talking to AI.
80% of cybersecurity is white-labelling, claims a builder
O3 Security's Rohit spent ten years infiltrating big tech companies for bounties, then scaled AppSec and CloudSec at Groww before building. His claim: most vendors resell the same open-source components (he once took over one and was 'inside every company' for four hours, unnoticed). His wedge is context - deep integration across the stack to kill false positives - plus compliance with CERT-In and RBI mandates; he reports ~$80,000 billed in five months with large fintechs and banks.
Compress hardware iteration from two months to one day
Placestation's solo founder - a self-taught electronics engineer who picked his college because his dad worked there and the punishment for bunking was 'call your parents' - builds desktop pick-and-place machines and reflow ovens. Instead of a contract manufacturer demanding 5,000-unit minimums, a hardware startup can make 100-200 units at a desk, test the market, and iterate at software speed. His larger point: India produces few hardware engineers because the ROI per student favors software.
Nikhil's playbook: storytelling, negotiation, and self-permission
His advice to the cohort: take storytelling seriously (he hired an Israeli speech coach who makes him write for three hours without devices) because your one-line description is the same line you'll sell with; learn negotiation because 'people don't often disagree with ideas, they disagree with people' - know who wants to agree with you before you sit down; and stop living by other people's timetables: 'society has wrongly coined hypocrisy as a bad word.' His own capital allocation after a year of sector work: energy transition - solar, batteries, EVs, grid efficiency - because AI compute needs power regardless.
The Conversation Breakdown
Cohort 3: from 3,000 applications to eight startups
⏱ Untimed topicAfter a hands-on Goa bootcamp (Gallup strength-ranking sessions, breakthrough exercises, stone-paper-scissors), Nikhil hosts the eight selected startups at home with his dog Grace and sets the format: each founder explains what they built and the trend they bet on - and later, the room votes on which company deserves a hypothetical $100.
Prava Payments: letting AI agents spend money
⏱ Untimed topicSushant (Raebareli, BHU economics) and Shubham (who left a Japanese startup, and yes, built the AI girlfriend for tier-2/3 India that made Rs 50-60K a month) explain why AI assistants stop at discovery: the Perplexity side and the merchant side run different payment gateways that can't talk. Prava is the PSP-agnostic middle layer - integrate once, pay any merchant - launching in the US and Southeast Asia, with UPI ambitions for India.
OncoALERT and the pathology OS: two bets on Indian diagnostics
⏱ Untimed topicDr. Jayanti's needleless, ~Rs 150 oral-cancer biosensor is engineered for places with no electricity, labs or trained staff, sold through hospitals and insurance. The doctor-engineer-rocket-scientist trio digitizes pathology slides for diagnostic labs, runs AI on live microscope feeds, and maps biomarkers to targeted cancer drug therapies - capabilities today limited to ultra-elite tier-1 centres. Forced to choose, the room bets on the software company: kits are hard to distribute, software isn't.
antimattr: the Android for earbuds
⏱ Untimed topicSridipto and Sirsho, childhood friends, argue voice is becoming the primary interface and screens the secondary one. Their AI layer for earbuds - and eventually glasses and other wearables - pipes notifications, food orders and cab bookings into your ear. Nikhil probes the obvious threat: couldn't the phone, or Meta's Ray-Bans, just do this?
O3 Security: context over checklists
⏱ Untimed topicRohit, ten years of bug bounties and two and a half years scaling Groww's AppSec and CloudSec, calls most of his own industry a scam of white-labelled open source. O3's difference, he says, is context from deep stack integration - knowing that 100 engineers and 12 managers touched a feature - which kills false positives, plus compliance plumbing for CERT-In and RBI mandates. He reports ~$80K billed in five months with large fintechs and banks.
BloomRehab and Placestation: hardware for bodies and factories
⏱ Untimed topicPrashanth and Abhishek's fridge-sized rehab robot imitates a physiotherapist's motion for stroke patients at home - aiming to lift daily rehab from 45 minutes toward the three-hour gold standard, pending CDSCO licensing. The solo founder of Placestation, self-taught after bunking his dad's own college, sells desktop pick-and-place machines and ovens so hardware startups can build 100 units instead of 5,000 and iterate in a day instead of two months.
Aeyi, then Nikhil unplugged
⏱ Untimed topicSahil's Aeyi gives existing CCTVs contextual understanding - a fall in an empty corridor triggers an alert with image and context, born from his co-founder's grandmother falling while he was in Bombay. Then the conversation opens up: get a storytelling coach and write for three hours without devices; negotiate against people, not ideas; don't marry on your family's timetable ('I probably have 34 New Years left'); Jungian shadow traits and therapy; why hypocrisy is just change; legacy is for others, longevity for himself; trading can't be modelled because emotion can't; and why his team's sector research landed on energy transition.
Core Frameworks & Ideas
Design for 'if nothing else works'
OncoALERT's constraint-first method: assume no electricity, no lab, no trained staff - then build the product that still works. Population-scale interventions in India can't import North American delivery models; the constraints are the spec.
Distribution decides between good products
Given two credible cancer startups, the room bet on the one whose product was software: physical kits need pharmacies and cold chains, software rides infrastructure that already exists. When products are comparable, pick the one whose distribution is already built.
Know who wants to agree with you
Nikhil's negotiation rule: 'people don't often disagree with ideas, they disagree with people.' Before negotiating, figure out whether the other side arrived wanting to agree or disagree with you, then calibrate - project circumspection to the skeptic, vision to the believer.
Actionable Next Steps
- ✓Say what your company does in one line to someone outside your field this week; if they can't repeat it back, rewrite the line - it's the same line you'll sell with.
- ✓Take your product's hardest environment (no power, no budget, no expertise) and list which of its assumptions break; the ones that survive are your real product.
- ✓Before your next negotiation, write down whether the other side walked in wanting to agree or disagree with you, and adjust your opening to that person, not your deck.
- ✓Block three device-free hours with a blank page and write anything - Nikhil's speech coach's drill; storytelling is a trainable skill, not a talent.
Notable Quotes
“people don't often disagree with ideas, they disagree with people.”
“if you can't appeal to me in that one line, tomorrow, when you sell it, that's the same line you're gonna use.”
“I think legacy is for others, longevity is for myself.”
The One Thing to Remember
Distribution and storytelling beat invention alone: the founders who win can say what they do in one line, and the infrastructure that removes friction (payments for AI agents, one-day hardware iteration, software instead of kits) scales further than the cleverest product.