What I Learned Building Products Without a Product Team
In the last three years, I've shipped six working products.
None of them went through a sprint. None of them had a product manager. None of them required an engineering ticket. I built all of them myself — as a marketing leader, in the gaps between everything else.
This is not a post about AI replacing engineers. It's about a shift in how I think about the relationship between having an idea and shipping a thing — and what that shift has changed about how I work.
The Old Mental Model
Before I started building, my mental model of product development looked like this:
Idea → Spec → Engineering backlog → Prioritization → Sprint → Build → Review → Launch
The gap between idea and shipped thing was measured in weeks, sometimes months. And because that gap was so long, I was selective about which ideas I brought forward.
Most ideas — the ones that were useful but not urgent, the ones that solved operational problems but didn't move core metrics — I simply didn't raise. The cost of raising them was too high relative to the probability they'd get built.
What this meant in practice: a lot of problems I could clearly see went unsolved, because the path to solving them was too long and uncertain.
What Changed
The first thing I built was the middleware validation layer for a sale campaign pre-booking flow.
The problem: We needed to verify whether a website visitor was an existing registered user before showing them a pre-booking form. The LMS API contained this information, but exposing it directly to a landing page created security and complexity problems.
The solution I built: A Google Sheets and Apps Script layer that sat between the landing page and the LMS API. The landing page called my middleware, my middleware called the LMS API, and returned only the validation response needed.
- No sensitive data exposed
- No engineering ticket needed
- 75,000+ validations processed for a single campaign
I didn't build this because I had a plan to become a builder. I built it because waiting for engineering would have meant missing the campaign window.
That experience reset my mental model. The question stopped being "can we build this?" and became "can I build this?"
The Six Things I Built
1. Telegram Mentorship Automation
Stack: Google Sheets · Apps Script · Telegram API · WebEngage Status: Live · 5,000+ users · 200 channels
200 private Telegram channels across 50+ subjects. On purchase, the system identifies the subject, maps to the correct channel, generates a single-use invite link, and delivers it via WebEngage.
Lifecycle management:
- Subscription expired → access removed
- EMI default → access restricted
- Payment cleared → access reinstated
Zero engineering dependency.
2. 4DX Productivity Dashboard
Stack: Firestore · Firebase Auth · Next.js · Vercel Status: Live · 12 team members
A team execution tracker built on Firestore and deployed on Vercel. Tracks streaks, daily tasks, weekly goals, monthly metrics, and comparative analytics.
This was the project that taught me how to structure a proper application — authentication, database schema, client-server separation.
3. AI Smart Analysis Mentor
Stack: Custom GPT · MCP Server · OpenAI API · Testbook LMS API Status: Live · 10,000 users · ~5 min avg session
A custom GPT with MCP server integration that provides personalised test performance analysis — weak areas, score improvement paths, 7-day and 30-day prep roadmaps.
Built entirely without product or engineering support.
4. Current Affairs Video Generator
Stack: Python · OpenAI · ElevenLabs · Google APIs Status: Live
A Python pipeline that takes a quiz ID and language as input and produces a complete short-form video in approximately 5 minutes.
Automated pipeline:
- Script generation via OpenAI
- Voiceover via ElevenLabs
- Image fetching via Google APIs
- Scene assembly and merge via Python
Two formats: daily roundup and single news explainer.
5. AI News Aggregator and Chatbot
Stack: OpenAI API · Google Cloud TTS · RSS · Python Status: Live
Fetches OpenAI research publications and Google News daily, filters by date, summarizes per card via LLM, narrates via Google Cloud TTS, and surfaces an embedded chatbot that answers queries based on filtered news.
Built primarily to stay current on AI developments without manual curation.
6. Habit Tracker with Social Layer
Stack: Firebase · React · Vercel Status: Live · Personal use
Started as a simple timer and check-in tracker. Evolved into something significantly more complex:
- Friend requests and discovery
- Shared streaks and activity visibility
- Unique invite links
- Email and phone-based friend discovery
- Notification systems
The social layer emerged from a genuine need: accountability works better when it's mutual and visible.
What Building Taught Me About Marketing
Three things that changed how I approach marketing work:
1. Constraints clarify thinking. When you're building something yourself, every decision must be made explicitly — data structure, API design, error handling, user flow. There's no room for ambiguity. This discipline transferred back to how I think about campaigns and growth initiatives.
2. The gap between idea and shipped thing determines which ideas get tried. When that gap is months, you're selective. When it's days or hours, you experiment more freely. Some of the most useful things I've built were ideas I wouldn't have raised in a sprint planning meeting — too small, too operational to prioritize. Building them myself meant they got done.
3. Engineering thinking and marketing thinking are more similar than they appear. Both are fundamentally about solving a problem for a specific user. The tools are different. The core question — what does this person need, and what's the minimum thing I can build that delivers it — is the same.
The One Thing I'd Tell Someone Starting
Don't start with a complex project. Start with a problem you have today that you can clearly describe in one sentence:
"I need to know which of the 200 people on this list are existing registered users before I show them a form."
"I need a way to send each student a unique Telegram invite link within 30 seconds of their purchase."
"I need a dashboard that shows me my team's daily task completion in one view."
These are buildable. They're also genuinely useful — which means you'll actually finish them, because they're solving a real problem.
The habit of building, once started, compounds. Each project teaches you something that makes the next one faster. The mental model shifts from "I need someone to build this" to "I need to figure out how to build this."
That shift, more than any specific technical skill, is what's been most valuable.
Dhairya Mehta is Director of Growth and Marketing at Testbook, where he owns the P&L of India's largest exam-prep subscription. He writes about growth, India's developing market, and building things.
Post a Comment