Siddhanth Jain

On this Independence Day, we are proud to announce the launch of Vizuara's first-ever AI product Our product team (Pritam Kudale and Pranavodhayam P R) has been working on this for months, and today we're launching DynaRoute - an intelligent router that automatically selects the best language model based on your query complexity and category. Here is the problem we solved: Most applications use one expensive model (like GPT-4) for everything. Simple questions like "What is the capital of France?" get the same treatment as complex reasoning tasks. It's like using a hammer for every job, even when you just need a screwdriver. DynaRoute analyzes each query and routes it to the optimal model. Simple questions go to efficient models like Gemini Flash. Complex reasoning tasks get routed to more powerful models. The result? 50-70% cost savings while maintaining performance. We tested this extensively. On our GPQA benchmark, we achieve a 9.7:1 performance-to-price ratio compared to 6.9:1 for single model solutions. For typical enterprise queries, we're seeing 93% cost savings over 300k tokens. What makes this different from AWS or Azure routers is the full spectrum access. We support open source and closed source models across GCP, AWS, and Azure. Their routers lock you into their ecosystem - Microsoft only gives you GPT models, Amazon only Claude/Llama. The integration is seamless. Drop-in replacement for OpenAI API, works with CrewAI agents, includes an MCP server for Claude, and comes with a comprehensive dashboard showing real-time savings. This is Vizuara's first AI product for enterprises, and Raj Abhijit Dandekar, Rajat Dandekar, and I are incredibly proud of what the team built. We have transitioned from education to solving real industry problems. Try it at https://lnkd.in/dGSx2KRq - free tier includes 50k tokens to test the $$ savings yourself. We just posted our product launch video on Vizuara's YouTube channel, where we give you a very detailed product walkthrough. We also explain how you can use this as an individual customer or as an enterprise to save $$$: https://lnkd.in/dtEHtHMn We will soon be publishing a paper on the concept of dynamic intelligent routing. Stay tuned. Happy Independence Day to my fellow countrymen! 🇮🇳

279

20 Comments

Bangalore is India's most overhyped startup city. I have been analyzing exit data across Indian cities, and the numbers surprised me. Bangalore receives 47% of India's startup funding. But generates only 31% of successful exits. Meanwhile, Delhi NCR shows a 0.67 cost-to-success ratio compared to Bangalore's 1.52. What does this mean? Every rupee invested in Delhi produces nearly 2x more exit value than in Bangalore. Here's what the data reveals: → Average operational costs 3x higher in Bangalore → Talent costs inflated by 40-60% due to competition → Infrastructure challenges are reducing productivity by 25% → Burn rates 200% higher than tier-2 cities Even established unicorns are rethinking their location strategy. BlackBuck recently relocated operations, citing 90-minute commutes affecting team productivity. This isn't about Bangalore vs other cities. It's about efficiency and sustainability. Some interesting patterns I have noticed: Tier-2 cities produce 26% of startups with 500+ employees despite receiving only 7% of total funding. That's remarkable efficiency. Companies in Pune, Hyderabad, and Chennai are building lean, profitable businesses from day one. Lower costs mean longer runways. Longer runways mean better decision-making. Better decisions lead to sustainable exits. The opportunity here isn't to abandon Bangalore. It's to recognize that great startups can emerge from anywhere. Founders in tier-2 cities often have advantages: → 2-3x longer runways with the same capital → Less competition for talent → Stronger focus on unit economics → Government support and incentives Maybe it's time we looked beyond the traditional startup hubs. Innovation doesn't need a Bangalore pin code. It needs the right problem, the right team, and the right execution. Curious to hear from founders across different cities, what's been your experience with costs and growth? #StartupEcosystem #IndianStartups #VentureCapital #Bangalore

432

112 Comments

BITS Hyderabad startup delivers combat drones to the Indian Army   Jayant Khatri and Sourya Choudhury, straight out of BITS Pilani, Hyderabad Campus Hyderabad, started just two months ago.   And in that short span, they’ve already delivered fully indigenous UAVs to Army units across Jammu, Arunachal, Panagarh and Chandimandir.   These aren’t concept prototypes or test trials. These are fully operational, combat-ready drones now being used on the ground.   Built from the ground up, these UAVs are designed to suit India’s diverse terrains and unpredictable field conditions.   What’s even more impressive is how simple they are to operate. Soldiers with no prior flying experience are being trained in just a few days, a huge step in making drone tech more accessible and mission-ready.   Their systems support the Army with real-time surveillance, long-range tracking, tactical payload drops and kamikaze missions. One of their high-speed kamikaze platforms can hit speeds over 300 km/h and deliver a 1 kg payload with precision.   That turns it into a powerful frontline tool that helps reduce human risk in critical missions.   The duo even showcased their technology recently in front of Kumar Mangalam Birla at the BITS campus. A proud moment for any student-led startup.   But more than just a moment, this feels like a shift. A reminder that student innovation can directly contribute to national security and solve problems that actually matter. #startup #innovation #bits

1,419

15 Comments

India’s thriving used vehicle ecosystem just got a major boost. Vutto, a fast-growing used two-wheeler marketplace, has raised ₹61.4 crore in #funding as it aims to redefine trust, transparency, and affordability in India’s pre-owned mobility sector. #startup https://lnkd.in/gXH9Q-X5

100

4 Comments

BIG TECH SHIFT -Mukesh Ambani just announced Reliance Intelligence - a brand new AI subsidiary of Reliance Industries Limited Reliance Industries launches AI unit 'Reliance Intelligence' Backed by Google & Meta, it will build world-class AI infra, deliver services in health, edu, agri & SMBs & train talent.

156

12 Comments

If you had to build a hypothetical, fast-moving, AI-backed startup to shake up a traditional industry ruled by incumbents, which one would you target, and how would you do it? That’s the challenge we’re throwing to smart, ambitious college students across India. I’m super excited and thrilled to launch the second edition of The Ken's Case Competition, in partnership with Zerodha. Last year’s debut was built on a format that was unique and powerful. Our intention was to radically reimagine the tired, old, formulaic case-study format to simulate the real world, with all its choices, complexities, and challenges. And so, in 2024, at the peak of India’s 10-minute delivery boom, we challenged India’s top MBA students to pick a real-world quick-commerce company and craft a winning strategy. The premise? We gave all teams a notional $1 billion in capital to spend as they saw fit. Teams from hundreds of B-schools went head-to-head over a three-month period in a fiercely contested, innovative battle for the winning prize of Rs 10 lakh—making it the case competition with the biggest prize in India. This year, we’re taking it up a notch. In most case competitions, students step into the shoes of a company and design a theoretical strategy to tackle a business problem. Our competition inverts this. The Ken’s next case-study challenge casts students as disruptors. They’ll analyse incumbents across sectors, uncover blind spots, design attack strategies, and eventually build AI-powered prototypes that show how they’d turn opportunities into products. Thus, we’re calling this a case-build competition—a first-of-its-kind, truly pioneering format. We’re doing this because the world has changed. It’s time for students to do much more than “study” stiff and constructed cases. Instead, it’s time for them to build out their true ambitions and ideas. Against incumbents in the real world. Because AI is forcing a reset of everything we knew about innovation and disruption. In the age of AI, the biggest opportunities are in identifying where legacy incumbents fall short. They’re holding onto customers whose needs have changed but don’t yet realize it. If AI can make anyone a builder, why can’t students be builders too? The winner of The Ken’s Case Competition 2025 will be awarded a cash prize of Rs 10 lakh. We envision this to be the gold standard of business and technology competitions in India. Open to all college students across the country, it’ll take place over three months, lasting until the end of 2025. In each round, participants will be pushed to go deeper into the sector of their choice—they’ll evaluate what incumbents are doing, conduct market research, sharpen their approach, and eventually craft and build solutions, using AI tools to bring them to life. And just like last year, we’ll involve The Ken’s community of subscribers by publishing the best and most innovative solutions and feedback we receive. Link to case in comments.

129

6 Comments

Can AI livestreamers finally make livestreaming work in India?   For years, livestreaming has been unsolved in India. Despite being time-rich with a social-first internet, livestreaming as ecom format never took off like China. It wasn’t for lack of trying - dozens of players entered but the pieces never quite clicked   There are two key issues 1) Supply side never scaled. High quality hosts with skill to entertain/ sell "live" are incredibly hard. Even when you find them, time ROI doesn’t justify effort. W/o massive concurrent views, even great hosts can’t monetize their time well 2) Demand side never hit critical mass. Watching live video to shop isn’t a habit for most users. Unlike China, where buying from livestreams is normalized (even aspirational), we haven’t built that instinct   It is one of the hardest content formats to crack. It’s not just about entertainment - it’s about holding attention for long stretches, establishing emotional trust, selling well. That’s a high bar.   China, however, cracked the code.   In China, livestreaming isn’t just content - it’s an industry. There’s a whole ecosystem of training schools that turn novice into conversion machines. There’s trust in format, amplified by exclusive SKUs + deep discounts. And now, something magical is unfolding..   AI livestreamers Luo Yonghao, China’s most prominent livestreamer, recently launched his AI avatar on Baidu’s ecom with their foundational model. Results were incredible 13M+ views $7.6M in GMV 6 hours of seamless, emotionally engaging, product rich conversation "LIVE" by AI avatar   It delivered 97k+ character of product storytelling, 8k+ coordinated gestures, even held banter with co-host avatar, complete with sync movement. It mimicked spontaneity, emotional shifts, tonal variety - everything that was once thought to be uniquely human. What’s more interesting: it did better sales than human host’s original debut   Why this matters I believed livestreaming was one content format AI wouldn’t touch. Too human, too spontaneous, too dependent on emotional resonance. That assumption’s been shaken.   AI livestreamers now deliver: 24/7 performance w/o burnout Cost-efficiency (<$1K setup vs a full-fledged studio) Hyper-consistent quality in storytelling, gestures They may not replace top-tier influencers, who thrive on emotional intimacy & narrative originality. But they could easily replace the long tail - those who drive volume, not creativity   Can livestreaming finally work in India? AI livestreamers could solve the supply side issue. With avatars that can shoot 24/7, mimic human emotions, deliver crisp storytelling at very low cost. Could this unlock ecom in catg like food, electronics, impulse-led lifestyle where trust equation is lighter & trial isn’t mandatory? It’s too early to say. But India’s livestreaming moment might not come from human creators It might come from AI avatars - fluent in emotion, optimized for conversion & built to scale where humans couldn’t.

56

9 Comments

We built an AI voice model for India - for just $6 🤯 Over the weekend we fine-tuned a LLaMA 3B model using audio as language on our proprietary dataset of 135 hours of speech across 9 Indian languages: Hindi, Tamil, Telugu, Marathi, Kannada, Malayalam, Punjabi, Gujarati, and Bengali. The model can not only convert text to speech and perform code-switching, but also clone voices across multiple Indian languages. For example, a native Hindi speaker can be made to speak in Tamil or any other Indian language. 👉 Try it out (https://lnkd.in/gyYr6SkU).  Ask it to generate audio in your native language or even mix multiple languages just for fun! 📝 Check out the blog (https://lnkd.in/gZn4gxEk). See how it compares with commercial models like ElevenLabs, Sarvam, and smallest.ai, as well as open-source models like Orpheus. The blog also includes technical details about data processing and the training process, along with code, datasets, and models that anyone can freely use and build upon. We firmly believe that just as anyone can now use AI to generate images, videos, or products, they will soon be able to build their own AI models to do almost anything - the only real bottleneck will be data. If you're building AI models and believe data is the key bottleneck to AGI, let’s connect. DM me, I'd love to talk about how we can work together to push the frontier toward AGI. Huge shoutout to Hugging Face, Unsloth AI, vLLM, Modal, and especially Canopy Labs🙌 - for the base model, Orpheus. Video credits: Jeyasurya U R — he cooked this video on a very short notice!

2,595

231 Comments

While the world watches the AI battle between Google and OpenAI, a new titan quietly entering the ring right here from India Mukesh Ambani's Reliance is building a dedicated AI empire for India. We're talking about "Bharat GPT," an AI that understands all our vernacular languages, and "JioFrames," smart glasses with real-time translation capabilities. The real game-changers are the companies backing Reliance. With giants like Google, Meta, and NVIDIA in his corner, Ambani isn't competing with the giants, instead he is working with them. The plan is to is to build a foundational AI ecosystem tailored for over a billion Indian users. This could be a defining moment, positioning India as the next AI powerhouse. India is soon emerging as the top largest market for AI in the world, and we might soon have some of the biggest developments in AI come out of India. Is the country ready for an AI that truly speaks our language? Let us know in the comments and follow Full Disclosure for more.

136

2 Comments

Natasha Malpani Launches ₹200 Crore Fund to Power India’s AI Startups Former Kae Capital partner Natasha Malpani has launched Boundless Ventures, an early-stage venture fund with a corpus of ₹200 crore, dedicated to supporting homegrown AI-native startups across India. The fund will focus on pre-seed and seed-stage investments in companies innovating across consumer AI, infrastructure, agent tooling, healthcare, logistics, deep tech, and make-in-India hardware. Boundless has already written its first cheques to startups including Superhealth, Armatrix, PierSight, Knot, and two stealth ventures in AI infrastructure and consumer AI. Beyond capital, the fund promises to bring narrative and network support to founders. As Malpani puts it, India today has a once-in-a-generation advantage talent density, digital rails, and ambition to build global AI leaders from day one. At Startup Veda, we’re thrilled to see local capital fueling the AI frontier. This fund signals rising maturity in India’s venture ecosystem—placing conviction behind AI-first founders well before the rest of the world catches on. Join the conversation: Which early-stage AI ventures excite you the most? How powerful is the mix of capital + storytelling for building AI startups? #Startup #AI #VentureCapital #DeepTech #StartupIndia #FounderStories

230

12 Comments