◆ Publications · Index
Writing on product, AI, and data.
A collection of LinkedIn publications on AI product strategy, growth experiments, and the craft of building data-driven features.
◆ Featured · Pinned
Editor's pick
Building a RAG-based Mutual Fund FAQ Chatbot for HDFC MF
NextLeap Learn in Public Challenge (2/4). A friend asked me a simple question — "What's the lock-in on my ELSS fund?" — and spent 20 minutes lost in a 168-page Scheme Information Document before giving up. So I built an AI assistant that answers factual questions about 4 HDFC MF schemes (Large Cap · Flexi Cap · ELSS Tax Saver · Mid-Cap Opportunities) with a source citation on every answer. Stack: vanilla JS · BM25 retrieval over 30 curated chunks · 20 official sources (HDFC AMC, AMFI, SEBI, CAMS, KFintech) · advisory-intent refusal layer · multi-chunk comparison for cross-scheme questions. #NextLeapPMFellowship #RAG #FinTech #BuildInPublic
App Review Insights Analyser for Groww
NextLeap Learn in Public Challenge (3/4). A PM friend at a fintech told me: "We get 200+ app reviews a week. I've never read a single one." So I built an AI analyzer that turns 120 public Play Store + App Store reviews from the last 8 weeks into a one-page note in under 10 seconds — top 3 themes, 3 verbatim user quotes, 3 action ideas, and a one-click email draft to stakeholders, auto-sent every Monday 9 AM. Stack: React + Tailwind + shadcn/ui · Supabase (Postgres + Edge Functions + pg_cron) · Resend API · 5 theme clusters (Performance, Support, KYC, Payments, Statements) + Positive bucket · medoid quote selection with verbatim validator. #NextLeapPMFellowship #ProductManagement #AI #FinTech
All Articles
Sorted: latest ▾
Introducing BookMyTrip — AI Agent Trippy
Last month, I watched a friend spend 6 HOURS planning a 5-day trip — 15+ Chrome tabs, flight prices changing every minute, hotel reviews contradicting each other, "best places to visit" blogs all sounding the same — and he still wasn't sure if he booked the right trip. Travel planning today is broken; we don't enjoy planning anymore, we just manage chaos. So I built BookMyTrip — AI Agent Trippy, a multi-agent travel planning assistant that thinks like an entire travel team working for you 24/7. One agent finds the best flights, another compares hotels, another builds personalized itineraries, another optimizes budgets, and another discovers hidden local experiences based on your travel style. You just say: "Plan my 4-day Bali trip under ₹60K with good cafés, nightlife, and peaceful stays" — and Trippy handles the rest. The future of travel isn't just booking tickets; it's eliminating decision fatigue. #AI #ArtificialIntelligence #TravelTech #AIAgents #GenerativeAI #Startup #BuildInPublic
Introducing Plately.AI — An AI-Powered Restaurant Recommendation Platform
Last weekend, my friends and I spent almost 45 minutes deciding where to eat — one wanted biryani, another wanted a peaceful café vibe, someone else cared only about ratings. We still ended up at a disappointing place. In 2026, we can use AI to generate videos, write code, and automate workflows… but finding the right restaurant still feels unnecessarily difficult — sponsored listings, generic recommendations, endless scrolling. So I built Plately.AI — an AI-powered restaurant recommendation platform that suggests places based on your mood, preferences, budget, and dining intent (not just random listings). 1,556 restaurants across India in V1. Would genuinely love your feedback. #AI #ProductManagement #Startup #FoodTech #ArtificialIntelligence #BuildInPublic #UXDesign #Innovation
AI PRD Generator with Multi-Agent System
Built an AI-powered PRD generator that turns a one-line product idea into a structured, high-quality "Product Requirement Document" in minutes. Instead of staring at a blank Google Doc, I assigned the work to a team of specialised AI agents — each with a clear role: PM Agent drafted the executive summary, Researcher mapped the user/market, Designer laid out the UX, Eng Lead outlined the technical architecture, Data Scientist defined the KPIs, GTM built the launch strategy, CS flagged support scenarios & edge cases, and DevOps designed the infrastructure. The agents collaborate, hand off context, and the output is a coherent PRD ready to review — not a generic AI brain-dump. #AI #ProductManagement #BuildInPublic
Make.com New User Onboarding Teardown
NextLeap Learn in Public Challenge (4/4). A founder friend pinged me: "My Zapier bill jumped from $30 to $400 last quarter — everyone says Make.com is 80% cheaper." So I tore down Make.com's full new-user onboarding — signup → first successful scenario → free-to-paid. A 12-slide deck covering business overview, 3 personas (Solo Founder, RevOps Lead, Curious Beginner), screen-by-screen PM rationale on all 9 onboarding screens, friction-mapped user journey, business→product outcome tree (Active Scenarios × Modules × Runs → Revenue), 5 prioritized PM recommendations, and a North Star + AARRR + onboarding-health KPI frame. Sources: Make Help Center, Academy, Community, G2 (1,800+ reviews), Zapier 2026 reviews. #NextLeapPMFellowship #ProductTeardown #SaaS #Onboarding
Data-Driven Decisions in Football and Product Management
NextLeap Learn in Public Challenge (1/4) — Moneyball Challenge. What if football clubs made decisions like Wall Street traders — purely by the numbers? I used SQL on a football dataset (players, teams, transfers) to find top performers by goals, best value-for-money picks (goals per salary), team performance summaries, transfer patterns, hidden underperformers, and team budget efficiency (wins per ₹ spent). Biggest takeaway: expensive ≠ effective. Players like Erling Haaland delivered the most goals at a fraction of the cost — exactly how product thinking works: optimise for value per unit of investment, not cost. #NextLeapPMFellowship #SQL #DataDrivenDecisions