Pallav Rustogi

It’s easy to critique a system. It’s much harder to build one. The recent nationwide delivery partner protests have sparked a heated debate: How does a platform like Zomato/Swiggy actually treat its fleet? Having spent six years at Zomato leading the tech teams for Logistics — specifically overseeing Rider Experience, Accounting, and Support — I’ve spent thousands of hours at the intersection of code and the gig economy. Here is what I learned about the complexities of building systems for a gig workforce at scale. Building a platform (Zomato | Blinkit | Swiggy | Zepto) that 9 lakh people use for 8+ hours a day is an immense challenge — especially when many of those users are navigating complex technology for the first time. It is easy to criticize a system; it is much harder to build one. These companies are constantly searching for a ‘sweet spot’ that doesn’t exist in nature: a world where consumers want free delivery, merchants want zero commissions, delivery partners want high earnings, and investors want profitability. These four forces pull in opposite directions every single day. My suggestion? Seek depth before forming a strong opinion. Understand the trade-offs, be respectful to the partners on the road, and acknowledge the effort of Indian companies trying to solve uniquely Indian problems at scale. if you are not a medium member, access the article using this link - https://lnkd.in/gSgVxCjg

454

17 Comments

Two weeks back I posted here seeking suggestions on Appxiom's elevator pitch. Thank you so much Sandeep Balaji, Hariprasad C, Arjun V Paul .., Nitin Vasanth and Shyam Sasidharan for sharing your insights and suggestions. We value that very much. The post triggered an offline conversation with Brajesh C. Kaimal PMP, DASSM, co-founder of Experion Technologies who visited us yesterday. We spent 90 minutes in intense discussions on Appxiom’s positioning. Both me and Don Peter kept wondering why we hadn’t spoken to him earlier about Appxiom, despite knowing him for years. It made me reflect on networking. What’s the point if it doesn’t create value? #Mentoring isn’t about teaching new things; it’s about helping us think beyond our silos. Combined with sharp product thinking, that’s what Brajesh brought to the table. Here is an example. Appxiom identifies #bugs and performance issues in applications. In traditional bug detection tools they are ordered on the basis of Recency, Frequency and Severity. But there is another additional factor - Business Impact. The sign-up flow with a http call delay will impact the business more than a crash in settings page. So what Appxiom does is to connect the bugs to the #userflows. Unlike analytical tools that tells you which user flows are impacted but not why, and unlike the bug detection tools that tells you which bugs occurred with out providing context and visibility into user journeys, Appxiom binds both data points. How to communicate this to the target audience in simple words? This is the crux of what Brajesh told us. While recency and frequency are tangible factors with absolute value representation, severity is more perception based and is subjective. I consider http call delay is severe because of the knowledge I gained through my engineering experience in the industry. For someone with strong UI background, a slight delay in screen loading could be more critical than the http delay. What Appxiom does is to redefine severity based on business impact and attach a score. So if the http delay is happening in signup page and the screen loading delay is happening in, say, settings page, http delay there is more severe. And if the screen loading issue is happening in signup page and http delay in settings page, screen loading issue takes priority. Appxiom makes sure the intangible factors in bug prioritizing are taken out and replaced with tangible data points enabling faster decisions, faster fixing and better quality. The pitch isn’t final, but we now have clarity. It’s no longer about bugs or user journeys. It’s about replacing intangible judgment with tangible, actionable insight. Thanks a ton Brajesh. Looking forward to more such interactions with you. The #startup ecosystem in #Trivandrum is richer because of the presence and involvement of people like you.

52

6 Comments

At the Amrit Niti roundtable, while contributing to the whitepaper that will be presented to the Government of India at the AI Impact Summit, one theme kept coming up: AI is not just eliminating jobs. It is redesigning the value chain of work. There’s a new role in tech right now paying $250k plus. Job postings up over 800 percent this year. Most engineers still have not clocked it. Forward Deployed Engineer. Companies like OpenAI, Anthropic, Salesforce and Stripe are hiring aggressively. I have included a few Job Posting in the first comment, do check them out. Why? Because AI does not fail at intelligence. It fails at implementation. Across industries, 70 to 80 percent of AI pilots do not deliver ROI. Not because the model is weak. Because every organization operates differently. Different workflows. Different compliance structures. Different risk frameworks. You cannot just plug in AI and expect transformation. A Forward Deployed Engineer sits between deep tech and business reality. They understand the model architecture. They sit with the enterprise. They customize workflows. They translate intelligence into impact. At the policy level, this is critical. If India wants AI driven financial empowerment and inclusive growth, we do not just need researchers. We need integrators. System architects. AI translators. People who can embed intelligence into real economic systems. Yes, certain roles will shrink. Repetitive analysis. Process heavy junior functions. Low leverage operational tasks. But entirely new high leverage roles are emerging. The shift is not AI versus humans. It is low context work versus high context work. The builders, integrators and system thinkers will win this decade. If you are in tech, stop asking whether AI will replace you. Start asking whether you understand it deeply enough to deploy it at scale. That is where the opportunity is. And if you are looking for a strong Forward Deployed Engineering capability to make your AI systems actually work inside complex enterprises, reach out to us at Navik Labs. That is exactly where we operate. #AIIndia #FutureOfWork #AIImpactSummit #Web3 #AIInfrastructure #DigitalTransformation

55

5 Comments

👥💔🏠How To Overcome Loneliness In The AI Era👥💔🏠 After intervening posts on Clicks Communicator, I am back to discuss loneliness as a serious mental health issue in the modern era. Never before have we been so connected and yet so lonely. We have apps to meet and chat with strangers, chatbots to resolve grievances, and AI for counselling. People are craving connection but dependant on tech/AI for emotional support. Loneliness is one of the biggest challenges, with 1 in 6 people worldwide affected by it, says the WHO. It has sparked a business model that’s monetising loneliness, as we saw in a previous post. Tech is being harnessed to develop useful solutions as well, including mental health apps and companion robots that engage in conversations. Johns Hopkins University students have even developed an AI-powered board game that combines NLP + gameplay to engage older adults. While such measures offer momentary relief, they are not a long-term solution. AI is programmed to be agreeable, but lacks empathy. It can end up encouraging people to self-harm or give incorrect advice, leading to disastrous consequences. People can also become addicted to apps (to the point of digital dementia), which can blur reality, and there are ethical/privacy concerns. Nothing can substitute for human interaction and lived experience. Perhaps that’s why we are seeing a resurgence of innovative offline groups. 🕺There’s "bhajan clubbing", which PM Narendra Modi has also spoken of — referring to groups of youngsters coming together to sing devotional songs in a blend of tradition and modernity. 🏃Fitness rave parties are also taking place across India where people meet to exercise/lift weights/go on a run, have protein shakes/energy drinks together, and dance to live DJ music. These are new-age versions of book clubs and hobby groups where people meet, bond over shared interests, and develop a community beyond work/family. 📲We even face isolation within families as we are caught up in our tech bubbles. To remedy this, in my family, we have adopted a non-negotiable level of interaction, and it’s helping. 👪Another concept that’s gaining traction is intergenerational living. For instance, in the UK, affordable housing schemes have elderly residents living alongside college students. In exchange for spending time/going on walks with the elderly, students are charged lower rent. In China, groups of middle-aged/elderly people are opting for shared housing to beat loneliness. In India, paid services like Aaji Care and The Family Member offer non-medical senior care. A volunteer visits to talk to a senior citizen, accompany them to the doctor/bank/shop or read to them and play board games. While there won’t be a one-size-fits-all solution, the way ahead is to focus on building communities over profit and acting as a catalyst for people to develop more real-world connections. #Loneliness #AI #Technology #Community #WHO 

18

2 Comments

Cursor built a product millions love. Tata Consultancy Services quietly out-earned them on AI. Here's the difference. To give you some detail - Cursor, the AI coding tool that hit $1B ARR in 24 mos did roughly $2B in ARR in Feb 2026. At the same time, according to a Concall, TCS crossed $2.3B in annualised AI revenue in Q4 FY26 alone. → A 57-year-old Indian IT company is neck and neck with Silicon Valley's hottest AI product. The reason is worth understanding. Cursor sells a tool. TCS sells transformation. Obviously the quality of revenue of Cursor is >TCS because the latter is service revenue. So its predictability is not long term. 📍 Cloud modernisation for enterprises that can't move without hand-holding 📍 Digital engineering for companies that have the budget but not the talent 📍 AI deployment for Fortune 500s whose internal teams are not yet AI-ready That last one is the key. I have seen this up close as we place AI talent into large enterprises and GCCs. The number one problem these companies have is not access to AI tools. They already have the subscriptions. The problem is readiness. Their data is messy. Their teams are not trained. Their infrastructure wasn't built for AI demands today. → Whoever solves the readiness gap wins. TCS understood this. That’s why their $2.3B is not from building AI. It is from making AI work inside an org that has: scale money and no patience for things that don't deploy cleanly. The next decade of AI revenue will not go to the companies that build the models. It will go to the companies that make enterprises actually use them. India is at the centre of that opportunity. Do you think we're training enough people to capture it? #ai #tcs #cursor

143

16 Comments

“22-Year-Old Founder Builds ₹100 Cr Startup.” “IIT Dropout Raises ₹80 Cr.” “This Startup Went From 0 to ₹50 Cr in 18 Months.” You have seen these headlines. What you usually do not see: • the burn rate   • the failed experiments   • the co-founder fights   • the customers who never returned   • the layoffs after the funding announcement   • the business losing money on every order  Startup media today is filled with numbers designed to look larger than reality: • Funding raised   • Valuation   • Downloads   • “Fastest growing”   • GMV   • User count  Most readers quietly translate those headlines into: “These people figured life out.” “They are winning.” “I am falling behind.” That is where the damage starts. A funding round is not proof of success. It is proof that investors are willing to take a bet. Those are not the same thing. A startup can raise ₹40 crore and still shut down. A company can process ₹100 crore worth of orders and still lose money. Growth alone does not mean the business works. The dangerous part of glorified startup coverage is not just bad reporting. It quietly pushes people into emotional comparison and uninformed risk-taking. Someone leaves a stable job because entrepreneurship now looks guaranteed. Someone burns savings chasing growth without understanding margins, retention, or cash flow. Someone mistakes visibility for business strength. Most startup failures are not mysterious. Usually the warning signs were already known: • weak pricing   • high customer acquisition costs   • low repeat customers   • co-founder conflicts   • growth dependent on discounts  The lessons existed. The headlines ignored them. Before starting a company: Study failed startups in your exact space. Not just the winners. Success stories create excitement. Failure stories create preparation. Big numbers attract attention. Understanding the business creates survival.

11

1 Comment

💡 Self-Reliant in Tech: The Time is Now The recent tariff discussions are not just headlines — they’re a reminder (and maybe even a disguised opportunity ) for us in India to rethink our dependency on global tech giants. For decades, we’ve leaned on: - Google for search & tools - Microsoft for productivity & operating systems - Amazon/GCP/Azure for cloud - Windows for… well, almost everything in offices -ChatGPT But here’s the big question: what happens to our economy and industries if these dependencies get disrupted tomorrow? Will it impact our ecommerce, online payment systems, essential online services, and more. 👉 The answer lies in building our own. 👉 The first step? Reforming how we teach — starting in schools, carried through to universities. 🎯 Encourage students (and professionals!) to experiment, fail, and try again. 🎯 Allow ingenuity and local innovation to shine. Imagine if even 10% of our graduates were trained to think and build this way. That’s enough to reduce, if not replace, the external dependencies we constantly worry about. Yes, it sounds ambitious. But so did building rockets in the backyard once upon a time. 🚀 And look where ISRO is today! So let’s get moving — experiment, create, and support a Made-in-India Tech ecosystem that can stand tall in the face of any global turbulence. The best time to start? Yesterday. The next best? Today. PS: If we start now, may be in a decade we will reach our goal. #MakeInIndia #TechLeadership #FutureReady

30

1 Comment

We wanted to test this on something that actually stresses systems. So we picked the Meesho app - huge surface area, tons of flows, real world complexity. And it ran beautifully. Introducing the full version of what we teased earlier: Parallel execution in Mahoraga by Quash with effectively no cap. Most platforms: “2 parallel devices”, “you hit your limit”, “upgrade for more”. Quash: run everything at once. No ceiling. Spin up hundreds of tests → everything fires in parallel Results → live, within seconds Testing that doesn’t wait for its turn. Divyesh Kiran Anand Sreejith Vijayendran Kumar Prasanna Arunachalam Navneet Chandra Alok Sharma Aditya Prasad M Feedback?

44

4 Comments

💰 One growing frustration among younger investors is this: A large portion of SIP money in India still flows into legacy businesses — 🥤 Consumer staples ⚡ Utilities 🍔 Food companies 🏦 Traditional conglomerates Many of these are stable businesses, but they are not necessarily where the next wave of exponential technology growth will come from. Meanwhile, globally: 🚀 AI ☁️ Cloud 🤖 Robotics 🧠 Semiconductors 📡 Deep-tech platforms …are creating the next generation of MAANG-scale companies. The problem? Most pure-play AI leaders remain private or are based overseas. That’s why more Indian investors are increasingly looking at: 🌍 US markets 📈 Global ETFs 🤖 Overseas AI firms …because they want exposure to innovation-led growth, not just defensive cash-flow businesses. There’s a strong case for SEBI to gradually open regulated pathways for retail participation in high-quality private tech and AI companies. Otherwise, domestic capital may continue flowing abroad in search of higher-growth opportunities. https://lnkd.in/dDfwbHy5

1

In my interactions over the last few days at the India AI Impact Summit, it seems like there is a great potential in integrating generative AI (RAG, agents, contextual recommendations, multi-modal / multi-lingual models etc) with ECHO's community-based learning model to further bridge the knowledge divide - however we must work to empower front line workers instead of replacing them, while making communities more resilient towards future pandemics and other crises. Impact Ignited- AI Powered Ideas in Action at InvokED 5.0 brought together insightful ideas from a diverse group, where I shared how Project ECHO is developing foundational technology to build and grow trusted human networks — amplifying the human impact of the 100 million+ front line workers globally A.I. is not a silver bullet that will magically solve challenges of inequity — it however can play a big role in democratizing access and empowering FLWs and eliminate barriers of region, language, and tech-literacy. This is the core of what our work at Project ECHO is all about.

166

3 Comments

What if I told you one of India’s most valuable tech companies… doesn’t even call itself a tech company? 🤯 It’s a government-owned railway portal that quietly became a ₹60,000 crore empire. Yes, I’m talking about IRCTC. Launched in 2002, it sold just 29 tickets on Day 1. Today, it books 13 lakh tickets a day, more than the entire airline industry of many countries. ✈️🚆 It’s India’s largest e-commerce engine by transaction volume, with 66 million+ registered users, alll from a platform that began with just one booking button. But what makes it dangerous is the tech. This is the same site that once crashed at 10AM every day. Today, it handles: → Handles 15,000 bookings/min → 3L+ concurrent users → AI chatbot “AskDISHA” (5 Cr+ queries, speaks Hindi + English) → Predictive analytics for demand surges → Real-time cloud infra that auto-scales & blocks bot attacks And the business model is insanely profitable with margins that would make most SaaS founders jealous. → ₹4,270 Cr in revenue → ₹1,111 Cr in net profit  → ₹16+ profit per user → Monopoly in online rail ticketing → Government-mandated convenience fees () CAC, infinite margin) → Cross-sells on every travel add-on All built in-house, at scale, without funding decks or founder AMAs & still 62.4% owned by the Government of India 🇮🇳 Now with Navratna status, IRCTC has the freedom to invest, expand, and even acquire, aiming to become a global OTA. We underestimate boring businesses but IRCTC proves that in India: Boring x Essential x Scalable = Unbeatable. You don’t need a new category. You need to own a daily ritual. And if you do it right, you won’t need to market. You’ll just be… inevitable.

1

Built my first MCP server. It finds offline businesses using Serper and sends the data to Clay via a webhook. After 6 months of deliberation. And multiple discussions with Anirudh and Romit, where I asked them 10000 questions, I finally built it. Why now? Because I feel it has become important enough that businesses can’t afford to ignore it anymore. Especially automation agencies like us. Rapid adoption from all three big LLM providers Anthropic, OpenAI and Google. Plus platforms like Zapier, Postman, Pipedream have made it possible to (i) easily create servers and/or (ii) integrate it into your workflows for a growing number of use cases everyday. What does my MCP server do? 1) It automatically creates the right queries to search on google using Serper. Selects the cities I want to target based on the states I mention. 2) Allows for batch creation and modularity for the number of pages I want to search. If I am looking for hotels I'll search for 20 pages vs searching for physical therapists where 5 pages should suffice. 3) Stores the API responses into a local folder as a csv file. Each run has a dedicated folder so I can always go back and access the data if I want. 4) Extracts the business details from the JSON responses and creates the final csv I can import to Clay. 5) Then it gets interesting because it also asks me if I'd like to directly send this data to a webhook table in Clay. This means even the manual import step is not necessary. For now at-least it seems like there are a few gaps: 1) Context window is a challenge, with Claude’s desktop client at-least. 2) Tool selection is not perfect. Both these things will drastically improve in the coming months. My assumption also is that using a client like Cursor will make things better, need to test. The big takeaway I had was the benefits of building your own server. I thought earlier, if out of the box MCP servers will become more readily available then users won’t need to spend as much time building, configuring or implementing their own servers. However, it turns out building your own server gives you the advantage of capturing your business logic more reliably. For example even if there are out of the box servers available for Serper (tongue twister much), I wouldn’t be able to achieve some parts of my workflow with it. Like my querying, batching logic, or the ability to transfer data via webhook to Clay. I could have potentially integrated other MCPs or tools to achieve the same but as I mentioned tool selection is currently a problem. There may be even better approaches to solve for a multi step workflow which captures your business context and needs - than just directly building that into your MCP server. Need to explore and build more.

75

20 Comments

JAMCO is expanding its global footprint with a new Product Engineering Centre in Bengaluru. This move strengthens India’s position as a growing hub for aerospace innovation, engineering excellence, and advanced manufacturing capabilities. A major boost for India’s aviation and aerospace ecosystem. Read more: https://lnkd.in/gpff8WRz #JAMCO #AerospaceEngineering #ProductEngineering #Bengaluru #AviationIndustry #EngineeringInnovation #MakeInIndia #AerospaceGrowth

4

Go Performance Tuning at Scale: Zepto’s pprof Journey (Part 1) At Zepto, performance isn’t just a priority — it’s an obsession. The best part about making systems more efficient is that it not only boosts speed but also drives real cost savings. In this blog, Akash Agarwal shares our journey of performance tuning — the challenges we tackled, the wins we unlocked, and how the process didn’t just improve our systems, but also shaped the way our engineers think about building scalable, efficient systems. Blog link: https://lnkd.in/gtzdtmGg #ZeptoTechXpress #PerformanceTuning Shivanshu Kumar Tiwari Ashish Bhatt Sumit Gupta Aditya Haridas Menon Harshal Khodifad Abhinav Nimesh Uday Solanki Devansh Chaudhary Devansh Goswami Sumit Ojha Vivek Gover Prashant Giri Balram Vishwakarma Shubham Prashar Anoop Singh Mayank Agarwal Nikhil Mittal

124

India just redefined AI-powered commerce. And it's not about chatbots. It's about trust, scale, and guardrails. On Oct 10 at Global Fintech Fest, NPCI, OpenAI, and Razorpay announced a collaboration that could change how 500M+ Indians discover and pay for products. But here's what excites me as someone building AI systems: This isn't just about AI agents processing payments. It's about building AI commerce infrastructure with: ✅ Compliance-first design - Built on India's trusted UPI rails ✅ Conversational commerce - Shop in your language, pay instantly ✅ Responsible AI at scale - Guardrails embedded from day one ✅ Financial inclusion - Bringing AI benefits to tier 2/3 cities The questions I'm asking: 🔹 How will AI agents verify user intent for high-value transactions? 🔹 What guardrails prevent fraud in conversational checkout flows? 🔹 How do we ensure AI doesn't push unnecessary purchases? 🔹 Can this work seamlessly across 22+ Indian languages? 🔹 What's the compliance framework for AI-driven financial decisions? Why this matters: We've seen AI content generation scale globally. But AI-powered commerce at India's scale—with built-in payment infrastructure and regulatory alignment—this is uncharted territory. If done right, this could be the blueprint for responsible AI commerce worldwide. My hope: That we build this with the same rigor we built UPI—user-centric, secure, and accessible to every Indian, not just the tech-savvy few. Kudos to Pragya Misra, Dilip Asbe, and the teams at National Payments Corporation Of India (NPCI), OpenAI, and Razorpay for taking this bold step. To fellow AI builders: How do we ensure AI commerce agents are helpful, not manipulative? Where should the guardrails be? This is the conversation India—and the world—needs to have. #AICommerce #ResponsibleAI #UPI #FintechInnovation #IndiaStack #AICompliance

6

Nice blog from Kailash, CTO of Zerodha TLDR- An experienced developer who can talk well, that is, imagine, articulate, define problem statements, architect and engineer, has a massive advantage over someone who cannot, more disproportionately than ever https://lnkd.in/gcKSkGuK

55

2 Comments

Folks working for big brands looking to adopt AI think that "understanding hindi,marathi is a solved problem now in 2026" Gemini has already solved the problem. But a simple experiment would be eye opening. Go to youtube and record a regional marathi interview from a farmer, tell gemini to transcribe. You will realize that the word error rate is considerable. To googles credit, Gemini 3 has definitely been an improvement over gemini 2. For example, the model gets "kenvha" and "kavaa" (Only rural marathi, not found in urban) mixed up. It is biased towards former. Former means "when" and later means "sometimes" and "when" depending on the context. A lot of marathi words will be found only in rural dialects and not at all in urban language. Another interesting thing to note is, google probably does not use these same models for transcribing youtube videos. The transcription are SO BAD. They get the language wrong and Word error rate is more than 80% for these regional audios. So search works pretty well for english videos because google does use transcripts but not good for regional videos missing very good descriptions. From time to time, I also like to watch models hallucinate even after telling them to fact check the answer. I took dialogues from a marathi movie and asked gemini if it recognizes the movie dialogue. Originally I wanted to do this to find out if google used movie transcripts for training. What I found was gemini confidently kept giving the wrong movie even after telling it to fact check. A human tends to say, "I don't know". The dialogues did contain names of the character. gemini just wouldnt verify if the character indeed exists in the movie. To be fair I did pick a movie released in 2013

11

The recent IT layoffs are revealing the harsh reality of the Indian Tech market. They are not just because of AI or a slowdown in the economy. Previously,  - One guy writes the frontend code. - Another guy writes the Backend code.  - Tester tests the software. - The product manager handles the stakeholders and clients. However, in the office, AI has now entered and is making work easier. If you encounter any issues, you can ask ChatGPT for the code. If you face challenges writing the test case, you can send screenshots to ChatGPT, and it can also provide you with the test case. Now, what is becoming more important? And I think all the founders will surely agree with this point.

4