Advertise on podcast: Z47 Moments

Rohan Joshi grew up in a Mumbai home where nobody said the word “entrepreneur,” but everybody lived it. His father builds pharmaceutical machinery, his mother runs a salon chain and her own cosmetics brand. So when his parents told him to “go get a job for experience,” he already knew what the experience was really for: a checkpoint on the way to building his own company. In this episode of Unstarted x Intelligent Indians, shot in the San Francisco Bay Area, Rohan sits down with Avneesh to talk about the story behind Stellar Labs (better known as KittenML), an AI research lab building open-source voice models so efficient they run on a CPU. They get into getting deep in AI before it was cool, computer vision at CMU, and three years at Meta building photorealistic avatars. Why he and his co-founder didn’t wait for permission to start building. The honest experiment that broke their first idea: pricing a local ElevenLabs alternative at full price and watching every user churn.  The pivot to developers, open-sourcing a voice model, and watching it go so viral on Reddit they assumed their own analytics were broken. His framework for knowing when to build instead of doing customer research. The three inflection points coming for voice AI. And the mindset that ties it all together: the easiest person to fool is yourself, so fool yourself into believing you can do anything. If you want to build in AI, this one is a playbook on conviction, shipping fast, open source, and why the five-year grind that used to separate the experts is now something AI hands you for free. 0:00 Intro: an AI lab that runs on a CPU1:20 Growing up in an entrepreneurial Mumbai family2:10 "Do a job for experience": the unspoken goal3:15 Getting into AI before it was cool (CMU, 2019)4:30 What actually changed: data, GPUs, and the transformer7:00 From medical AI to Meta's photorealistic avatars8:20 Starting Stellar Labs with a co-founder from Meta9:30 Why they didn't wait for permission to build10:40 Conviction vs delusion: should you do customer research?12:00 Why the biggest companies started before they were businesses13:30 The first idea: a local alternative to ElevenLabs15:30 The experiment that broke it: everyone churned16:40 Why open source opens a bigger market than closed18:20 How do you know it's working? When you doubt your own analytics20:40 The future of voice AI: three inflection points24:30 Compounding: what looks like talent is years of work26:40 Voice AI that runs on a $20 server28:00 Should you move to the Bay Area to build?29:50 Parting advice: fool yourself into believing you can

Shashank N.D. dropped out of college to build Practo, back when dropping out just meant you couldn’t get a job. No business family. Both parents in government service. When everyone around him said “do an MBA first, you can always start a company later,” he flipped the logic: you can always get a job later, but right now I have the least baggage I’ll ever have. So he stayed in, bootstrapped Practo, and sold software to doctors one month at a time, often unable to afford even an ice cream. In this episode of Unstarted, Shashank sits down with Avneesh, who has known him for 13 years and backed Practo through its highs and near-death lows. They talk about the story behind the story. Building Practo for his father, who had been through 7-8 surgeries. Finding product-market fit purely through doctor referrals. The frameworks that held the team together for 18 years. The brutal years when doctors burned his billboards and FIRs were filed for the crime of rating doctors. Why COVID turned Practo from a pariava into the industry. And the mindset that kept him going: fall in love with the problem, and measure impact, not valuation. If you’re building, or still unstarted, this one is full of hard-won lessons on conviction, co-founders, culture, surviving crises, and playing a decades-long game. Chapters 0:00 Intro: the oldest, youngest founder0:52 Bangalore roots and a government-service family1:58 Discovering entrepreneurship at the E-Cell5:50 Bootstrapping and the ice cream he couldn't afford6:49 Co-founders, trust, and 18 years together8:44 From stocks to legal to healthcare: finding the idea10:00 Two business cards: CEO for VCs, sales exec for doctors11:30 Why "Practo", and the $6 domain name11:51 Market timing: fall in love with the problem, not the timing15:43 The only two metrics that mattered for survival17:35 Missionary vs mercenary founders25:50 Culture: the WHOOPIM framework28:57 "Do Great" and leading by example30:47 His mother's advice that changed how he manages35:51 Death by capital: the 2015 funding mistake38:27 COVID: the pariah becomes the industry40:11 Profitability, the US launch, and AI in healthcare42:50 A note on the Matrix partnership43:54 Advice to the unstarted: measure impact, not valuation

Ajay grew up in Bangalore chasing free pizza at hackathons, trained as a machine-learning researcher, and never planned to start a company until his brother’s accident and a nightmare healthcare experience in New York pushed him to build Coral Health. In this episode of Unstarted by Z47, Ajay sits down with Anish to break down how Coral is killing traditional RPA (robotic process automation) and cutting US healthcare wait times from 14–25 days down to under 5 minutes. We go deep on workflow automation vs. “thin AI wrappers,” why the fax machine still runs American healthcare in 2026, how to tell real customer feedback from polite feedback (The Mom Test), pricing on value (the 10x rule), closing your first customer on the way to the airport, and the co-founder he recruited 5 minutes before a VC meeting. If you’re a technologist wondering how to turn deep expertise into a real business, this one’s for you. 🎧 Unstarted is the podcast for founders, by founders - from Z47, India’s oldest venture capital firm. 00:00  kill RPA, care in under 5 minutes01:00   San Francisco & meet Ajay of Coral Health01:45  What Coral Health does: two missions03:00  Born in California, raised in Bangalore04:30  Free pizza at hackathons: the love for tech08:00  The brother's accident & stages of grief10:00  Falling in love with the RPA problem by accident13:00  The friend's gallbladder & the broken patient journey16:00  Stress-testing the idea like a scientist18:30  $5–7 to maintain every $1 of RPA24:00  Closing the first customer on the way to the airport26:00  The Mom Test: real feedback vs. polite feedback31:30  Why healthcare still runs on fax machines in 202635:00  Why OCR fails: 55% accuracy on 720-page faxes42:00  Why this is deep workflow automation, not a thin wrapper44:00  Co-founders: the Aniket story (recruited in 5 minutes)51:00  Advice to the unstarted: find a problem, solve it, have fun

Chetan Reddy, founder of Confido, joins the Intelligent Indians podcast to explain how AI is leapfrogging one of the least digitized industries on earth: US healthcare an industry where, as he puts it, the system of record is still sticky notes. Confido builds AI teammates agents with names like Sarah and Ryan  that run the front office of specialty medical practices: answering phones, handling insurance, booking patients, chasing referrals and payments.  The result is striking. Around 800 to 900 patients a day ask to meet or speak to Sarah, not realizing she's AI. Fewer than 2% of callers now choose to speak to a human up from a 65% AI-completion rate when they started to 98% today. In this conversation Chetan and the Z47 team go deep on why healthcare resisted tech for so long, why "digitizing the work" has to come before any AI ROI, how Confido built and humanized its agents with 2,000+ empathy parameters and constant patient feedback, how their on-site deployed engineers (FDEs) and the "Care OS" platform make each deployment faster than the last, the India (Bangalore) team that powers 24/7 patient coverage, the open-source vs frontier-model strategy, and the bigger vision: becoming the operating standard for US specialty care. Intelligent Indians is brought to you by Z47. What you'll learn: 1. Why "digitize the work" must come before any AI ROI in a legacy industry2. How to humanize AI agents so customers prefer them to humans3. Why fewer than 2% of patients now choose a human over the AI4. How on-site deployed engineers + a platform make each deployment compound5. Why an AI-native company can beat foundation-model giants on last-mile ROI6. Open-source vs frontier models: a practical framework for vertical AI Chapters 00:00 Intro -  how US healthcare runs on 1990s software (and sticky notes)01:30 Why AI can leapfrog the digitization friction04:00 "Can I meet Sarah?" - 800 patients a day asking for an AI06:00 Humanizing agents: task completion, 2,500 empathy parameters, patient feedback09:30 Why less than 2% of patients choose a human11:00 Can't a foundation model just do this? Why digitizing the work is the moat14:00 What "digitize the work" actually means — the 25-insurance-contracts example16:30 Landing deployed engineers inside a care org without conflict18:30 The outcomes: 2–3x more patient encounters, 20–30% cost savings20:00 Building Care OS — a platform whose customer is the FDE23:00 How deployments compound: 6–9 months down to 3–625:00 Is the market as competitive as it looks?29:00 The Bangalore team and 24/7 patient coverage33:00 The bigger vision: becoming the operating standard for US healthcare

Aniket Shah quit his job with no startup idea, gave himself 18 months, and ended up building Swish one of India’s fastest-growing 10-minute food delivery companies, now serving 1 million lifetime users with 220+ freshly made items. In this episode of Unstarted by Z47, Aniket sits down with Anish to unpack the journey from Ahmedabad to IIT Kharagpur, early stints at Citibank, Jupiter, Pillow Fund (crypto) and Meesho, and the 4 PM hungry-in-a-café moment that sparked Swish.  We go deep on product-market fit, why Pillow Fund shut down despite 1M+ users, choosing co-founders, the “digital-first QSR” thesis, kitchen automation, and why the best product not the first mover wins in food. Whether you’re a founder, operator, or just someone who orders a samosa at midnight, this one’s for you. 🎧 Unstarted is the podcast for founders, by founders, ⏱️ Chapters below 👇 00:00  The thesis: why 10-minute food should exist01:15 Crypto to samosas, meet Aniket Shah02:30  Ahmedabad childhood & unusual parenting04:00  Memory talent: reverse alphabet, cricket stats, WWE cards06:15  IIT Kharagpur & the entrepreneurial pull07:45  Citibank internship & turning down the offer09:30  Joining Jupiter: UPI product from day one11:30  Pillow Fund: the crypto 18% interest bet13:45  Why Pillow shut down — user love & product-market fit17:00  How to know if you have PMF18:30  A brief stint at Meesho19:30  Do you need a co-founder? Chemistry over physics23:00  Do you need an original idea?25:00  The 4 PM café moment that sparked Swish28:00  Owning the full supply chain & "neighborhood kitchens"30:30  The digital-first QSR thesis32:30  Kitchen automation, AI & consistency at scale36:00  Counterintuitive learnings & customer food habits39:30  Hits & misses: snacks, beverages, meals42:00  Competition: "run against yourself"44:00  Mumbai launch & why not expand to 5 cities at once46:30  Advice to the unstarted: don't wait for perfect48:30  Memory challenge & sign-off

Sumit Chhazed has started four companies. The first three didn't work. Solar heating, where he was too young to win government tenders. An edtech platform built with IIT Bombay professors that was years ahead of its market and ended in a modest acquisition. Then a two-wheeler commerce business born out of a second-hand Pulsar that broke down every third day in Bombay. The fourth is OTO Capital, a lending infrastructure platform for banks and NBFCs in two-wheeler financing. In 2024, mid-growth and mid-fundraise, the round fell through. He and his co-founders stopped taking a salary for twelve months, let people go, and rebuilt the company around profitability instead of growth. They turned down buyout offers along the way. In this episode of Unstarted, Sumit talks to Avnish Bajaj about how to tell when a business isn't working, building for yourself versus building for an investor, what changes when you switch a company from growth mode to profitability, how to communicate with a team through that shift, and separating conviction from delusion. Unstarted is a series for founders, by founders, with questions crowdsourced from the audience. Chapters 0:00 Intro: three startups that didn't work0:47 What OTO Capital actually does2:42 Why this business works now and not ten years ago4:48 Udaipur, a family of teachers, and the business keeda6:26 Selling chewing gum stickers in 6th standard7:51 Telling his parents he wouldn't take a job9:41 Startup one: solar, and poor founder-market fit12:45 Startup two: CourseWave, and being too early19:48 Q: How do you know when it's working?22:44 Q: Can you build without funding?26:17 The 2024 roadblock27:12 Choosing default alive over growth27:56 Letting people go, and how to do it right31:13 No salary for 12 months33:28 Why he turned down the buyout offers36:10 The 2001 Kolkata Test37:02 Closing: it's a Test, not a T20

Srikanth Velamakanni built an AI company before most people knew the A in AI. In 2000, long before ChatGPT, he and five co-founders started Fractal Analytics with a simple bet: that math could predict human behaviour and help companies make better decisions. The market wasn't ready. HDFC took 18 months to convince. Citibank made them prove themselves against 100 in-house PhDs. They won on accuracy, stayed profitable, and 26 years later Fractal became India's first AI company to go public. In this episode of Unstarted, Srikanth sits down with Avnish to talk about the story behind the story. Growing up middle class in a small oil town in Assam. Why he never thought he was founder material, and the Narayana Murthy lecture that changed his mind. If you are building, or still unstarted, this one is full of hard-won lessons on conviction, timing, co-founders, and staying relevant across 26 years of technological change. Chapers 0:00 Intro1:15 From a small town in Assam to Fractal2:47 Math plus human behaviour: an AI company before AI3:34 "I never thought I would be an entrepreneur"5:01 The Narayana Murthy moment5:54 Not money: building a 100 year company6:57 Conviction vs market readiness9:32 When the market is not ready (ICICI, HDFC)10:27 The Citibank bake-off against 100 PhDs12:00 Compounding your core passion15:44 The two co-founder types you actually need18:39 Finding your ICP without chasing noise19:16 26 years of AI, from regression to transformers23:08 Strategy is what you refuse to do26:24 Elected CEO, then asked to leave in 6 months28:02 Defining the company's values after crisis29:19 Raising money and investor expectations31:34 Will AI make Fractal obsolete?33:58 Why Fractal was built in Mumbai36:30 Advice: build what OpenAI cannot kill

Amit Lakhotia moved from a village in Churu, Rajasthan to Delhi at 10, speaking Marwari and barely any Hindi. Twenty people shared three rooms. His father couldn't afford the school fees, so he made a bet with the principal: if his son didn't come first within a year, throw him out. That year Amit scored 19 out of 20 in Maths and Hindi, and 1 out of 20 in English. His tuition was waived from class four through twelve. He went on to NSIT and IIM Ahmedabad, helped scale MakeMyTrip, Paytm and Tokopedia, built a taxi business that was four times Ola's size before shutting it down, and then founded Park+ — which raised its first cheque two months before COVID shut down every mall and office in the country. In this episode of Unstarted, Amit talks to Avnish Bajaj about when to walk away from a working business, the vitamin versus painkiller test for choosing what to build, why he picked parking as the wedge into the car ownership market, how access control emerged from the worst possible timing, and why brand and revenue aren't a trade-off if you think creatively. Unstarted is a series for founders, by founders, with questions crowdsourced from the audience. Guest: Amit Lakhotia, Founder & CEO, Park+Host: Avnish Bajaj, Founder & Managing Director, Z47 Chapters 0:00 Intro: the founder who gets away with scolding me1:14 Churu to Delhi, and 20 people in three rooms2:09 Arriving at school with no English3:04 The bet his father made with the principal5:01 How he survived that first year5:47 19 out of 20 in Maths. 1 out of 20 in English.7:21 NSIT, IIM-A, and where the confidence came from9:20 Sitting in his father's shop and deciding to drop out10:12 Turning around the IIM-A souvenir store12:15 GetMeCab: profitable in month three13:42 Why he walked away from a business 4x Ola's size14:50 Q: I want to start a business. How?17:57 Vitamins vs painkillers18:38 The two ideas he shortlisted, and why parking won21:24 Q: How do you measure progress during a pivot?21:36 Two months in, COVID arrives22:52 Turning a pandemic into the access control business24:38 The metric that mattered: client acceptance25:35 Revenue vs brand, and getting signage for free27:43 Closing: build the network, and give before you take

Nikita Khanna spent fifteen years straightening her hair before she realised the problem wasn't her hair. It was that nobody had built products for it. She was on partner track at McKinsey her only job ever when she started going down YouTube rabbit holes during COVID, trying to find a sustainable way to manage wavy, frizz-prone hair. She found the answers in Brazil: similar hair texture, dozens of local brands, none of it available in India. When she asked people in the industry whether it was a business, almost everyone told her the category was too small and too random. Roughly 80% of Indians have hair somewhere on the wavy-to-curly spectrum. She built Moxie anyway. In this episode of Unstarted, Nikita talks to Avnish Bajaj about validating an idea before you quit your job, the ABC framework she now uses to evaluate any new category, why she never thought she was cut out to be a founder, learning to delegate and build an org for the first time, running a company with her husband as co-founder, and what it actually takes to build a brand in a market where launching a product has never been easier. Unstarted is a series for founders, by founders, with questions crowdsourced from the audience. Chapters  0:00 Intro: the company we should have invested in00:00 Growing up in Delhi, watching her father build a business00:00 DU, XLRI, and landing at McKinsey00:00 Straightening her hair since 13, and deciding to stop00:00 Finding the answer in Brazil00:00 "Am I even cut out to be a founder?"00:00 Q: How do I validate an idea before I quit?00:00 The ABC framework: assortment, business, consumer00:00 Everyone said the category was too small00:00 Q: How does a young founder become a CEO?00:00 Building with her husband as co-founder00:00 Why viral products aren't the goal

Aman Goel, founder of Greylabs.ai, joins Avnish on Unstarted for one of the most relatable founder journeys yet — a middle-class kid from Kanpur who was hustling for money from his first year of college and never stopped. Aman is refreshingly honest that his motivation was simply to make money. He earned his first ₹4,000 solving JEE papers, built a content-writing business with his now-wife and co-founder Harshita, and had saved close to ₹50 lakh before he even graduated.  He went on to build Cogno AI - an AI chatbot and voice company for banks - in 2017, before generative AI existed, bootstrapped it to a million-dollar revenue run rate, and sold it to Exotel at 25. Now with Greylabs.ai, he's building voice AI for India's banks and financial institutions.  In this episode he breaks down finding customers in an industry you have no network in, the "student not salesperson" playbook that got him into SBI and HDFC, why he and Harshita split the company cleanly into "she builds, I sell," how to part ways with a co-founder without a fight, picking a razor-sharp ICP, and why in a world of weekly AI launches trust is still the only thing that sells. Chapters 00:00 Intro — a genuinely relatable founder story01:00 Growing up middle-class in Kanpur02:50 "My motivation was money" — where ambition came from04:00 The internet, Silicon Valley, and deciding not to settle in the US05:50 Earning in college: ₹4,000 to ₹50 lakh before graduating08:20 How to create your own opportunity in college11:50 Starting Cogno AI — and the RED vs REL framework15:40 Building AI in 2017, before ChatGPT17:20 Getting into banks: the "student, not salesperson" playbook21:40 Negotiating with SBI as two 20-year-olds24:40 The first customer, and the referrals that followed25:30 The co-founder exit: quick decision, generous separation27:20 Harshita joins: "she builds, I sell"30:40 The exit to Exotel — doubling the ask in 10 minutes34:40 Finding the market first: why he stayed in India37:50 Founder-market fit and picking a razor-sharp ICP40:40 Can AI disrupt his AI business? Why trust wins42:50 Eating your own dog food: building internal tools on Replit44:40 AI adoption in banks — from pilots to 50 lakh calls a month46:14 The coming pricing pressure, and how service is the moat49:20 Closing advice: persistence is the one thing that can't be replicated

Sameer Nigam, founder and CEO of PhonePe, joins Unstarted for a candid conversation on building a company that now serves over 700 million users. From a middle-class naval family to a failed music startup (Mind360), an acquisition by Flipkart, and the founding of PhonePe, Sameer breaks down the thinking behind his "asymmetric bets," why he chose UPI when the entire market was chasing wallets, and how the Big Billion Day payments collapse revealed the problem he'd spend the next decade solving. He also opens up on his 31-year partnership with co-founder Rahul Chari, why he embraces regulators instead of fighting them, what PhonePe's data reveals about how India really spends, and the story behind the IPO that was pulled at the last minute. A must-watch for founders thinking about co-founders, market timing, building for population scale, and knowing when to cut your losses. Unstarted is a show about founders, for founders hosted by Avnish Bajaj.  Chapters 0:00 Intro: meet Sameer Nigam & the "celebrity founder" tag2:15 Family background: a Navy dad and an entrepreneur mom at early TCS10:45 Why tech fundamentals still win (and the Google "75% AI code" point)17:40 First venture: Mind360, the "iTunes for India"22:10 Why it failed: wrong market sizing and the piracy era26:00 Acquired by Flipkart and the B2B pivot34:15 Big Billion Day: when India's payment infra collapsed38:00 Why wallets were never the answer41:20 The asymmetric bet on UPI44:30 Ghar Wapsi: merging back into Flipkart to buy time47:10 Demonetization: luck, timing, and near-zero competition50:00 Obsessing over infrastructure and the right foundation52:30 ICP, population scale, and the "painkiller in the hinterland"55:40 Incumbents push back: ICICI blocking and the RBI visit59:00 What PhonePe's data reveals about how India spends1:05:00 The most anticipated IPO that didn't happen1:08:30 Final advice: don't be an entrepreneur by FOMO

Saahil Goel, co-founder and CEO of Shiprocket, joins us for a candid deep-dive into how a self-taught coder from Delhi built the logistics and enablement backbone powering a huge chunk of India's direct-to-consumer commerce. Sahil takes us all the way back selling his dot matrix printer on Baazee as a kid, teaching himself HTML by hacking "view source," and building websites for cash in college. Along the way he shares the hard-won lessons behind every pivot: why Indian SMEs won't pay upfront, why you should never fall in love with your product, and how obsession with the merchant's problem kept the company alive through years when every investor said the market didn't exist. We also get into co-founder dynamics and trust, what actually changed scaling from 25 to 1,500 people, why culture is the "uncodified personality" of a company set by the founder's actions, and being a "microsurgeon, not a micromanager."  Finally, Saahil lays out his conviction on AI - from writing 70%+ of code with AI, to Shiprocket's in-house voice model built for India's 22 languages, to the "AI for Bharat" mission of reaching the next 58 million merchants. A rich, practical episode for anyone still unstarted. What they talk about: 1. From selling a printer on Baazee and self-taught coding to building Shiprocket into India's commerce shipping backbone 2. The real story behind the pivots  KartRocket to Kraftly to Shiprocket and why you stay stubborn with the problem, not the product 3. What PMF actually feels like, plus co-founders, trust, and scaling culture from 25 to 1,500 people 4. His bet on AI: writing most of the code with AI, an in-house voice model for India's 22 languages, and the mission to reach 60M MSMEs 00:00 Intro - Saahil Goel and Shiprocket02:30 Delhi boy, self-taught coder & selling a printer on Baazee07:00 Business family & knowing early he'd be a founder12:00 Meeting Gautam & the start of Bigfoot Retail Solutions17:00 Co-founders: trust, complementary skills & full commitment23:00 PMF is a journey, not a destination29:00 Stubborn with the problem, not the product35:00 CartRocket to Craftly: the pivots and hard lessons42:00 Building Shiprocket & why B2B behaved like consumer48:00 Scaling 25 to 1,500: culture, micro-surgery & builder mindset55:00 Bias to action vs bias for excellence60:00 AI at Shiprocket: 70% of code, voice for Bharat & 60M MSMEs70:00 The IPO ahead & advice for the still-unstarted

Our first-ever episode with co-founders. Karan and Himanshu of WootzWork join Avnish to talk about building a hardware and manufacturing exports company out of India — and why they leaned into a market everyone told them wouldn't work. From an investment banking and Stanford GSB background (Karan) and IIT Kharagpur mechanical engineering (Himanshu), the two met at a previous company where one interviewed the other as head of engineering a one-hour conversation that ran three hours. In this episode they unpack how a single insult ("we don't talk to Indian engineers, we work with Germans") lit the fire, how they closed their first order from India with no factory of their own, and why the naysayers became their strongest signal to go ahead. What you'll learn in this episode: 1. Why starting in a market everyone doubts can be an advantage — and how to tell real naysaying from noise2. How to win your first customer with no factory, no brand, and no track record3. The co-founder decision that actually matters: why the answer is never "the middle"4. How to choose a co-founder — honesty and complementary skills over everything else5. Why customer obsession beats worrying about network or capital early on6. Manufacturing as a system: how fragmentation held India back and how to fix it7. Scaling global sales by hiring domain experts in their 40s–50s who are local to the market Chapters 00:00 Intro 02:00 Karan's path: banking, AirAsia India, Stanford GSB04:30 Himanshu's path: IIT Kharagpur, ITC, and the machines05:15 "We don't talk to Indian engineers" — the moment that lit the fire06:30 The China vs India buying experience: what's actually broken08:00 Why the market was ripe — and betting on exports11:45 Customer obsession: getting the first customer in13:15 The co-founder dynamic: why the answer is never in the middle20:25 Turning the naysaying into conviction22:30 The first order: won with no factory23:12 Why the customer still needs them: engineering vs manufacturing28:30 Managing people with 40–50 years of experience30:30 AI and robotics in the manufacturing story33:40 Closing: why 2.3 is the right number of co-founders

What happens if you fail with everyone's money? Abhishek Bansal never let himself ask and he thinks that's the only reason Shadowfax exists.He's the first public-company CEO on Unstarted: a middle-class kid from Meerut who quit his job without telling his parents, asked a VC for $5 million with zero orders, and was the only one still standing after the year every competitor died.In this conversation with Avnish Bajaj, they attempt to answer: 1. Should you have a co-founder and how do you actually pick one? 2. What should founders look for in a VC when there are multiple offers? 3. How do you tell your team you're failing?4. Sustainable growth, or aggressive scaling? By the end, you understand his test: accept the failure, name the core problem, tell your team. Because teams don't break on failure — they break on how you behave in it. Chapters 0:00 Intro: our first public company CEO1:19 Meerut, a middle-class home, and IIT Delhi4:04 Parents didn't know he had quit his job7:50 Q: How do you choose between VC offers?8:47 Why he wanted a co-founder, not a skills match11:03 Why not bootstrap?11:28 Asking for $5 million with no product12:23 The ₹8 lakh phone call13:28 The angels who never sold: 1000x and counting14:31 Three offers: $2M, $5M, $9M17:19 Q: How do founders bounce back from failure?17:46 The 2015 crash: five funded, one survivor23:07 Knowing the core problem is the hardest problem28:07 The offsite: profitable in three months28:32 The retention numbers behind the culture29:20 Q: Aggressive scaling or sustainable growth?29:56 "In India, no idea is unique"32:14 The guardrails: unit economics and 50-30-2037:26 The CAPEX model that ran 100 iterations39:07 When the HR team built its own hiring tool40:20 Closing: teams compound, individuals don't

Most people assume serious AI silicon can only be built by a giant, in the cloud, burning hundreds of millions of dollars. Ravi Annavajhala did the opposite. Kinara built two AI chips for under $50M, 90% of it out of Hyderabad, and sold to NXP, one of the world's largest semiconductor companies for $307M,  all cash. It's one of the largest deep tech exits India has seen, and one of the least talked about. In this episode, Vikram Vaidyanathan sits down with Ravi to take apart how it actually happened, and what it says about where AI is heading next. You'll hear: 1. Why the disruption playbook that minted the NAND-flash exits is the same one that explains Kinara's sale to NXP 2. The clearest plain-language explanation of training vs. inference you'll find, and why inference is moving off the cloud and onto the edge3. The four reasons edge beats cloud: cost, latency, privacy, reliability 4. How you actually shrink a frontier model to run on a chip: distillation, compression, purpose-built retraining 5. Why India's real AI edge isn't talent or capital, but the industrial data sitting in its factories — and why it can't afford to become a "data colony" 6. Ravi's distinction between frugal and cheap, and why low cost has stopped meaning low capability ▶ If you're building, investing in, or just tracking India's deep tech story, this is the episode to watch. About Intelligent Indians Intelligent Indians is Z47's thesis podcast on the people building India's technology future, for builders, operators, and investors who want the operator's view, not the press-release version.🔔 Subscribe for new episodes. Chapters  0:00 A $307M deep-tech exit you didn't hear about0:45 Meet Ravi Annavajhala & the Kinara story3:10 The NAND flash lesson: how startups cash in on disruption5:45 2015: betting on edge AI while everyone chased the data center7:00 CoreViz → Deep Vision → Kinara: the naming journey9:15 Why 90% was built in Hyderabad — a talent bet, not a cost cut11:30 Two chips under $50M: engineering for capital efficiency15:20 What "the edge" actually is: training vs inference18:00 Why edge beats cloud: latency, cost, privacy & reliability22:45 Physical AI: robots, self-driving & the real bottlenecks28:30 Advice to VCs: conviction, local demand & the infra gap31:20 Frugal vs cheap: why low cost ≠ low capability33:30 Closing: India on the cusp of the edge-AI wave