Srishilesh P S

How I went from coding in hackathons to building PyroAI for 3.5 years, to founding 3 ventures, to becoming a Builder PM. The story of why I made the switch and what I learned along the way.

I didn't plan to become a Product Manager. In fact, I actively avoided it for years.

During my MBA at Masters' Union, while everyone was obsessing over PM case studies and frameworks, I was heads-down building Docupull, my AI-powered title insurance platform. "PM is too hand-wavy," I thought. "I'd rather ship code."

But here's the thing: the best products aren't built by engineers who code in isolation. They're built by people who understand WHAT to build and WHY.

This is the story of how I learned that the hard way.

The Hacker Origins (2017-2021)

My product journey started before I knew what "product" meant.

During undergrad at Amrita University, I participated in 25+ hackathons. Every single one felt like running a mini-startup:

Form an all-star team
Ideate on a problem worth solving
Build an MVP in 24-48 hours
Pitch to judges

In 2019, I won runner-up at an AI hackathon by building an LSTM model that separated pilot-ATC communications from noise. The next day, I jumped into ETHIndia (a blockchain hackathon) knowing nothing about Ethereum. Spent 36 hours learning and ideating.

That experience sparked an idea: fighting misinformation using blockchain and AI.

I cold-emailed Jimmy Wales (Wikipedia founder), got on calls with VCs like ColliderVC, joined the DataDAO community. Even the DAOStack founder was interested.

The result? Nothing shipped.

The technical complexity was beyond what I could execute as a solo founder. I had a vision but no path to make it real.

First lesson learned: Ideas are worthless without execution. And execution requires knowing WHAT is actually feasible and valuable.

The PyroAI Years: Building in Production (2021-2024)

I joined Mr.Cooper Group (NASDAQ: COOP) — the 2nd largest non-banking mortgage servicer in the US with 9K employees — in January 2021 as a Software Engineer Trainee building PyroAI, a B2B SaaS ML infrastructure for processing millions of mortgage documents.

Year 1: The Full-Stack Grind (Jan 2021 - Apr 2022)

Role: Software Engineer Trainee

Started building customer-facing features with React, Python, and Spring Boot. I was the "make it work" engineer:

Built PyroAI web & ML infra enabling extraction workflows for handling millions of documents
Implemented thumbnail generation service processing 1 million+ PDF images simultaneously with 98% extraction accuracy
Optimized PDF bursting service to efficiently handle 5M+ documents daily with 98% latency reduction

I was proud of my code. But I didn't understand the business.

I'd build features engineers loved but users barely touched. I optimized things that didn't matter. I spent weeks on technical elegance that added zero value to the end user.

Year 2: The ML Transition (May 2022 - Mar 2023)

Role: Software Engineer 1

After 1.5 years, I moved to the ML team as the 5th member. Youngest on the team.

Key projects:

Migrated ML model infrastructure, cutting infra costs by 50% and improving developer productivity by 30%
Built document clustering model in 3 weeks, reducing manual intervention by 50% while achieving 99% system uptime
Enhanced search UX with filtering capabilities, improving query relevance scores by 100% and retrieval time by 80%

But the bigger insight came from watching the operations team.

65+ people were spending 8+ hours a day manually annotating mortgage documents to train our models. Going through thousands of pages, clicking, labeling, repeat.

As an engineer, I thought: "Can't we automate this?"

Year 3: The Product Manager Moment (Apr 2023 - Mar 2024)

Role: Product Engineer 2 – Machine Learning

I proposed an initiative: build a Human-in-the-Loop (HITL) annotation tool that self-learns from previous annotations.

Instead of annotating from scratch, the model would pre-populate fields and annotators would just correct mistakes.

I led a 2-person team (2 MLEs) building the MLOps pipeline:

Auto-retrain on new batches
Update models in production
Deploy without downtime

6 months later, it was ready.

The Impact That Changed My Perspective

After a year in production:

Reduced TAT by 800% and labor costs by 50% for 65+ annotators
Millions saved in operational costs

But here's what hit me: This wasn't a technical win. It was a product win.

The hardest part wasn't coding the ML model. It was:

Convincing leadership it was worth investing in
Understanding which annotations to automate (not all were equal)
Designing a UX that annotators would actually adopt
Measuring impact in business terms (time saved, cost reduced)

I was doing product management without calling it that.

The Enterprise Search Project

My final project at Mr.Cooper confirmed the shift.

500+ call center agents were spending 8 minutes per customer call (costing $16+ per interaction) searching through 3000+ documents to answer questions.

I spearheaded user research with 500+ agents, identifying document management needs and defining the product roadmap.

I worked directly with an SVP as lead developer / business analyst / product manager (no one knew what to call me).

My role:

Built chatbot-like enterprise search interface
Served as primary liaison for Google Cloud Consulting on GenAI initiatives using VertexAI AgentBuilder

2 months later: MVP shipped, reducing AHT from 8 mins to 5 mins (-37.5%) with org-wide adoption.

Result: CEO's Wall of Fame recognition.

Realization: I spent more time understanding the problem, talking to users, and coordinating stakeholders than writing code.

And I loved it.

The MBA Years: Building 3 Ventures (2024)

By 2023, I knew I wanted to build products, but I lacked the business framework.

Joined Masters' Union for an MBA focused on startups. GPA: 3.6.

But I didn't just study — I built.

Venture #1: Docupull (Current)

AI-powered title verification platform that started before MBA and continues today.

Started as a B2C idea (autofilling mortgage loan applications from KYC docs) with a colleague who faced this pain on a student visa.

Launched on Product Hunt in Nov 2023: Top #3 Product of the Day.

Pivoted to B2B (title verification) in May 2024 after 50+ customer interviews.

Current state:

$800 MRR
800+ orders processed with 98-100% SLA adherence
100% repeat usage from paying clients

What I'm learning:

You need to talk to customers BEFORE building features
Pricing is product
Retention > Acquisition
I still code daily on Docupull (automations, integrations, bug fixes)

PM realization: The best PMs have founder DNA. They ship, iterate, and don't wait for permission.

Venture #2: Rugshak (Custom Rugs)

D2C brand selling custom rugs and carpets.

What I learned:

Performance marketing (Meta ads - achieved 3.26% CTR and 2.62 ROAS reaching 67K+ users)
High-ticket B2B sales (made ₹60K in a single day at an event with ₹20K AOV)
₹2.5L+ revenue in 3 months through B2B/B2C channels

Why I shut it down: Dropshipping is a grind. Margins are thin. No moat.

PM skill gained: Customer acquisition is harder than building features.

Venture #3: Bambaii Foods (Healthy Snacking)

D2C healthy snacks brand.

What I learned:

Product development (recipe iterations, packaging design)
Supply chain and inventory management
B2B partnerships (got into modern trade stores)
2,500+ units sold, ₹6.6L revenue in 3 months with 100%+ margin products

Why I shut it down: Food is operationally complex. I wanted to build software products with better unit economics.

PM skill gained: Operations matter. You can't product-manage your way out of a bad business model.

Vmart: PM at Scale (2024-2025)

Joined Vmart Retail (NSE: VMART) — an ₹800M NSE-listed value fashion retailer with 500+ stores and 10K+ employees.

Strategic Tech Initiatives – CEO's Office (Mar 2025 – Jun 2025)

Reported directly to the CTO conducting enterprise-wide process discovery and optimization.

Key initiatives:

Defined "Lost Sales" metric — Exposed 6% SKU-level sales leakage to guide assortment and inventory planning
Led org-wide research across 10 departments — Identified process gaps and prioritized 8 high-impact tech initiatives
Digitized competitor benchmarking — Eliminated 100-person manual surveys, reducing data collection cycle by 50%
Built CXO-ready operational reports — Data centralization reduced executive decision timelines by 50%

What I learned: Being Chief of Staff means being the "translator" between vision and execution. You need to diagnose problems before prescribing solutions.

Product Manager – Digital Transformation (Jun 2025 – Present)

Now owning the digital transformation roadmap across the entire retail chain with full autonomy.

Major wins:

Automated Open-to-Buy planning workflow — Reduced merchandising cycle from 15 man-days to 1 day per month
Piloted interstore transfer in 2 regions — Targeting 30% sell-through improvement and 50% logistics cost reduction
Built n8n automation framework — Created 5-6 workflows, saving 20-50 manhours per month per process to <1 hour
Shipped HR payroll automation for 10,000 employees — Cut payroll processing time by 90% across stores and warehouse
Created design-to-display fascia mapping tool — Piloted in 5 stores, improving planogram compliance accuracy to 95%

Impact: These automations are saving the company millions annually while dramatically improving operational efficiency.

Managing at scale is different than building from scratch. But my coding background helps me:

Call BS on impossible engineering timelines
Prototype fast instead of writing specs (built n8n workflows myself)
Debug production issues with the team
Speak both "business language" to CXOs and "tech language" to engineers

Why "Builder PM"?

Most PMs I've met fall into two camps:

1. Strategy PMs - Great at slides, roadmaps, stakeholder management. Can't read code.

2. Technical PMs - Understand systems, can write SQL, review PRs. But too deep in the weeds to see the bigger picture.

I'm neither.

I'm a Builder PM:

I code daily (Python, React, TypeScript) - last project: yesterday on Docupull
I ship 0→1 products - Docupull, PyroAI features, Vmart automations
I understand business - ₹9L+ revenue across ventures taught me unit economics
I move fast - Hackathon mindset never left. MVP > perfect spec.

What excites me about Builder PM roles:

Ambiguity - Give me a vague problem, I'll ship something in 2 weeks
AI x Product - We're in the golden age of AI. I want to build products that actually use AI well (not just slap ChatGPT on everything)
Startups - I don't want to "manage roadmaps." I want to build things that matter.

The Aha Moment: Why PM > Pure Engineering

At Mr.Cooper, I built a beautiful clustering algorithm. Clean code. Optimized. Fast.

No one used it.

Why? Because I built what I thought was cool, not what users needed.

Later, I built the enterprise search tool. The code was messier. The architecture wasn't as elegant.

But it saved millions of dollars and got me on the CEO's Wall of Fame.

The difference? I spent 80% of my time understanding the problem and 20% coding the solution.

That's when I realized: The best engineers solve technical problems. The best PMs solve the right problems.

I'd rather solve the right problem than write perfect code for the wrong one.

What I'm Looking For Next

I'm actively seeking Builder PM or Chief of Staff roles at growth-stage startups (seed to Series B, 10+ team size).

What I bring:

Technical depth - I can code, architect, and debug production issues
Founder experience - Shipped 3 ventures, know how to move fast
0→1 track record - PyroAI automations, enterprise search, Docupull
Business acumen - MBA + real revenue from multiple ventures

What I'm optimizing for:

Learning - Want to work with founders who've built 0→1 before
Autonomy - Give me a problem, let me figure it out
Impact - Startups where my work directly moves metrics

Location: Bangalore, Chennai, Hyderabad preferred. Remote/hybrid works too.

Ideal problem spaces: AI/ML products, Fintech, SaaS, Developer Tools

The TL;DR

Started as a hackathon addict who loved building
Spent 3.5 years at Mr.Cooper building PyroAI, realized the hard problems aren't technical
Got my MBA while founding 3 ventures (₹9L+ total revenue) to learn business
Discovered I'm a Builder PM - technical enough to code, business-savvy enough to know what to build
Now looking for my next 0→1 challenge at a startup that's actually solving real problems with AI

If you're building something ambitious and need someone who can code, ship, and figure it out as they go — let's talk.

P.S. Still not sure what "Builder PM" means?

Builder PM = Engineer who learned WHAT to build + Founder who learned WHY it matters

That's me.