Building AI Products That Turn Complexity Into Scalable Systems

Product Manager | AI Builder | Entrepreneur

Download Resume
0+
Years Building Products
0
Products Launched
0K+
Users Impacted
0+
AI Systems Built

Operator depth across every layer of product

Founder accountability, production AI systems, and a decade of shipping — not slideware.

0+

Years of Product Management

End-to-end ownership: discovery, PRDs, roadmaps, launch, iteration — across B2B, B2B2C, and consumer.

0

Companies Founded

Founder & CEO of Zerton Engineering Services and Zerton Education Technologies. P&L-level accountability.

0+

AI Systems in Production

LLM content pipelines, RAG systems, multi-agent orchestration, evaluation frameworks, voice AI.

0+

SaaS Institutions Served

Multi-tenant B2B2C ERP + learning platform rolled out across 26+ engineering institutions, 42K+ users.

M.Tech

Education & Foundation

M.Tech Machine Design, B.E. Mechanical Engineering. Published composites research at L&T. Continuous upskilling in LLM engineering, evals, and agentic systems.

0+

Chess Puzzle Rating — Top 0.15%

Crossed 3000 in Chess.com puzzles — top 0.15% globally. The same pattern recognition, calculation under constraint, and many-moves-ahead planning I bring to product strategy.

From engineering rigor to AI product leadership

Five stages. Each one compounded into the next — and left a lesson that still ships in every product.

012014 — 2015

Engineer

L&T Heavy Engineering · Research Intern

Researched fiber-reinforced polymer composites under marine conditions. Published findings on the dynamic failure behavior of GFRP. Learned to instrument a system, stress it, and read the data honestly.

Lesson Learned

Systems fail at the interfaces, not the components. Products do too.

022015 — 2023

Startup Founder

Zerton Engineering Services · Founder & PM

Spotted institutions running operations on Excel and WhatsApp. Built a B2B2C ERP + learning platform from 0→1, grew it to 26+ institutions and 42K+ users across Maharashtra with a cross-functional team of 8.

Lesson Learned

Nobody buys software. They buy fewer Monday-morning fires.

032023

EdTech Builder

Zerton Education Technologies · CEO

Launched Listen2RE — an audio-first learning platform for UPSC/MPSC aspirants. Shipped a mobile web MVP in 8 weeks instead of a 6-month native app. 35K+ learners followed.

Lesson Learned

Habit design beats feature count. Consistency is the product.

042024

AI Product Builder

LLM Pipelines · RAG · Agents · Evals

Re-architected Listen2RE around an LLM content pipeline with LLM-as-a-Judge quality gates — cutting production effort 60%. Built RAG systems, multi-agent automations, and an AI PRD generator.

Lesson Learned

The AI is invisible. The value is the outcome.

05Now

AI Product Manager

Agentic Systems · Product Leadership

Operating at the intersection of product strategy and applied AI: scoping what LLMs can reliably do, designing the eval harness that proves it, and shipping products users return to daily.

Lesson Learned

Ship outcomes, not models. Measure everything that matters.

A living system, not a checklist

Six phases, one continuous loop. Every phase below is illustrated with a decision from a real shipped product.

Phase 01Discover

Live with the user's problem before touching a solution. Field interviews, JTBD framing, and observing behavior — not just asking about it.

In Practice

Interviewing working UPSC aspirants surfaced the real constraint: not motivation, but 45–90 wasted commute minutes a day. That insight became Listen2RE.

Products as documentaries — problem to outcome

Full product lifecycle, shown end-to-end: research, strategy, architecture, metrics, and the lessons that survived contact with users.

LIAI EDTECH · 0→1

Case Study 01AI EdTech · 0→1

Listen2RE

An AI-augmented audio learning platform that turns wasted commute hours into UPSC progress.

Role — CEO & AI Product LeadTimeline — 2023 — Present
Claude APILLM PipelinesLLM-as-a-JudgeTTSn8nMixpanelPWA
35K+
Learners served
71%
Session completion
60%
Production effort cut
8 wks
Idea to launch

Problem

UPSC/MPSC preparation takes years, the syllabus is enormous, and most aspirants hold full-time jobs. The standard solutions — books, coaching, YouTube — all demand screen-on, focused attention that working aspirants simply cannot give.

PAIN-01

Working aspirants commute 45–90 minutes daily in conditions where reading is impossible — that time produces zero progress.

PAIN-02

Existing audio content was low-quality YouTube: slow pacing, no structure, nothing designed for audio-native learning.

PAIN-03

Aspirants knew exactly what to study. The constraint was never knowledge of the syllabus — it was usable time.

Research

Field interviews with working aspirants aged 24–34, employed full-time, with 1–3 hours/day for prep. Mapped their actual day hour-by-hour instead of asking what features they wanted.

I know what I need to study. I just can't find the time to sit and study it.

Working MPSC aspirant, user interview
  • 01Commute time was the single largest block of recoverable learning inventory — and it was 100% unused.
  • 02Aspirants had tried podcasts and abandoned them: content read aloud from documents doesn't work for ears.
  • 03Decision fatigue was real — large content libraries caused skipped sessions, not more engagement.

Insights

Insight 1Passive commute hours are wasted learning inventory — the product is reclaiming time users already spend.

Insight 2Audio-native ≠ text-to-speech. Content designed for ears has different sentence structure, pacing, and signposting than content designed for eyes.

Insight 3For habit-forming behavior, consistency beats choice. One great daily session outperforms an infinite library.

Strategy

Position Listen2RE not as a podcast or TTS app, but as an AI content system that ingests dense UPSC material and produces structured, audio-native daily sessions matched to each learner's commute.

Product Requirements

  • Daily 15–25 min audio session, pushed at 6:30 AM, matched to the learner's current topic
  • LLM pipeline: topic extraction → concept summarization → audio-script formatting → key-term callouts
  • Quality gate: LLM-as-a-Judge pre-filter plus 10-minute human spot-check before publish
  • Engagement loop: session rating, streaks, and replay with topic override
  • Mobile-first web app with PWA caching for offline listening

Prioritization — RICE, weighted by habit impact

Every candidate feature scored on Reach × Impact × Confidence ÷ Effort, with a habit-impact multiplier. Features that strengthened the daily loop (push timing, streaks, voice quality) consistently out-scored content breadth. The library expansion everyone asked for scored lowest — and was cut.

PRD — North-star metric

Completed listening sessions per learner per week — not downloads, not signups.

PRD — Explicit non-goal

Listen2RE is not an on-demand podcast library. One daily session, sequenced for the learner.

PRD — Quality guardrail

No AI-generated session ships below an 85% LLM-as-a-Judge pass threshold. Human review on every flag.

Execution

Built the full content system end-to-end: ingestion, LLM processing on the Claude API, dual-layer quality review, voice synthesis, and CDN delivery — orchestrated with n8n, instrumented with Mixpanel.

System Architecture

01Source material — UPSC docs, current affairs, PYQs
02Ingestion + chunking (Python)
03LLM processing (Claude API) — extraction, summarization, audio-script formatting
04Quality gate — LLM-as-a-Judge + human spot-check
05Voice synthesis — TTS pipeline
06Platform + CDN delivery

User Journey

  1. 16:30 AM — daily session lands, matched to current topic
  2. 2Commute — 15–25 min structured audio, paced for listening
  3. 3Key-term recap — callouts reinforce retention at session end
  4. 4One-tap rating — feeds the quality loop
  5. 5Streak + progress — visible momentum across the syllabus

Key Decisions & Trade-offs

AI generates, human spot-checks — not the reverse

Trade-offOccasional quality misses a human-first flow would catch. Accepted because the user feedback loop surfaces issues fast — and production effort dropped 60%.

Daily push over on-demand library

Trade-offLess perceived user control. Offset with topic overrides and replay. Habit strength won: 71% completion.

Mobile web first, native app later

Trade-offNo offline listening at launch — the #1 feature request. Shipped 8 weeks instead of 6 months; PWA caching closed the gap in a later iteration.

Metrics

35,000+
Total learners served
71%
Avg. session completion rate
84%
Report it reclaims commute time
87%
LLM-as-a-Judge quality pass rate
13%
Human flag rate — down from 31%
40%
Cost per session cut over 6 months

Lessons

01

Personalization earlier. Topic-sequencing by individual progress should have been month 2, not month 8.

02

Community sooner. Learners wanted to discuss sessions — a simple thread per episode would have lifted retention for 14 months.

03

Evals are pre-launch infrastructure. Shipping without systematic LLM evaluation cost us early retention. Never again.

04

Voice quality was the real product. The voice-pipeline upgrade beat every feature shipped that quarter.

BEB2B2C SAAS · 0→1

Case Study 02B2B2C SaaS · 0→1

B2B2C ERP + Learning Platform

From Excel-and-WhatsApp chaos to a multi-tenant platform running 26+ engineering institutions.

Role — Founder & Product ManagerTimeline — 2015 — 2023
Multi-tenant SaaSERPLearning DeliveryAnalyticsAPI IntegrationsGTM
26+
Institutions onboarded
42K+
Users on platform
8
Cross-functional team led
0→1
To multi-campus rollout

Problem

Engineering institutions across Maharashtra ran admissions, attendance, fees, exams, and learning delivery on Excel sheets and WhatsApp groups. No single system of record, no analytics, and every process depended on one overworked clerk's memory.

PAIN-01

Administrators rebuilt the same student data in five disconnected spreadsheets — errors compounded every term.

PAIN-02

Faculty had no channel for structured learning delivery; students had no unified view of schedules, marks, or materials.

PAIN-03

Leadership flew blind: no institution-level analytics on attendance, fee collection, or academic outcomes.

Research

On-ground discovery: campus visits across Maharashtra, shadowing clerks through daily workflows, structured interviews with directors, admin staff, faculty, and students — three very different users of one system.

Don't show me a dashboard. Show me that admission season won't break my staff this year.

Institution director, discovery interview
  • 01The buyer (director) and the daily user (clerk, faculty) had completely different success criteria — adoption, not features, was the real product.
  • 02Institutions didn't want 'digital transformation'. They wanted fewer fires: fee reconciliation that closes, attendance that tallies, marksheets that don't bounce.
  • 03Change capacity was the binding constraint — staff could absorb one new module at a time, not a big-bang platform switch.

Insights

Insight 1In B2B2C, the user who never signed the cheque decides whether the product lives. Design for the clerk, sell to the director.

Insight 2Operational reliability is the wedge — analytics and learning delivery only matter after the boring workflows are bulletproof.

Insight 3Phased, champion-led rollout beats big-bang deployment in low-change-capacity organizations every time.

Strategy

Build a modular, multi-tenant platform that wins on operational reliability first, then expands into learning delivery and analytics — rolled out campus by campus through trained on-site champions.

Product Requirements

  • Multi-tenant core: each institution isolated, centrally upgradable
  • Module sequence: admissions → attendance → fees → exams → learning delivery → analytics
  • Role-based views for director, admin, faculty, student, and parent
  • On-ground champion training program per campus as part of onboarding
  • Engagement analytics to drive data-informed iteration and churn reduction

Prioritization — Adoption-weighted value vs. implementation cost

Features were scored on value to daily users × likelihood of actual adoption ÷ rollout cost. High-prestige features the buyer asked for (fancy dashboards) were deliberately sequenced after high-adoption workflow features the clerk needed — because churn follows the clerk, not the director.

PRD — North-star metric

Weekly active staff per campus — the leading indicator of renewal, two quarters out.

PRD — Explicit non-goal

No custom one-off builds per institution. Configuration over customization, or multi-tenancy dies.

PRD — Rollout guardrail

No new module ships to a campus until the previous module hits adoption thresholds with its staff.

Execution

Led a cross-functional team of 8 across engineering, design, content, and GTM through the full lifecycle: user research, PRDs, roadmap, MVP, pilot campuses, and multi-campus rollout with on-ground enablement.

System Architecture

01Campus onboarding + tenant provisioning
02Core ERP modules — admissions, attendance, fees, exams
03Learning delivery layer — materials, schedules, assessments
04Role-based access — director / admin / faculty / student / parent
05Engagement + operations analytics
06API integrations — payments, SMS, government reporting

User Journey

  1. 1Pilot — one campus, one module, success criteria agreed upfront
  2. 2Champion training — on-site power users own internal adoption
  3. 3Phased module rollout — next module unlocks on adoption thresholds
  4. 4Feedback cycles — monthly campus reviews feed the roadmap
  5. 5Renewal + expansion — adoption data makes the renewal case itself

Key Decisions & Trade-offs

B2B2C model over pure B2B licensing

Trade-offHeavier support load serving students and parents directly — but it made the platform sticky at every layer of the institution.

Phased per-campus rollout over big-bang deployment

Trade-offSlower revenue recognition and longer sales cycles. Worth it: adoption survived staff turnover and exam-season stress.

Sales-led GTM with on-ground enablement

Trade-offDidn't scale like product-led growth — but in this market, trust is built in the staff room, not in a free trial.

Metrics

26+
Institutions onboarded and retained
42,000+
Students, faculty & staff on platform
8
Person cross-functional team led
6
Module categories shipped
5
User roles served per tenant
8 yrs
Operated and grown as founder

Lessons

01

The buyer and the user are different people with different definitions of success. Ship for both, in the right order.

02

Service-heavy onboarding felt like a tax — it was actually the moat. Competitors who shipped software without enablement churned out.

03

Data-informed iteration cycles reduced churn more than any single feature: campuses that saw their own adoption data renewed.

04

Configuration over customization is an existential rule for multi-tenant SaaS, not a preference.

An innovation lab of working AI systems

Not concepts — systems that run. Each card is something built, shipped, and measured.

AI Agents

Shipped

Multi-agent orchestration with n8n + LLMs: retrieval, reasoning, and generation agents chained with zero manual handoffs.

n8n·Prompt Chaining·Orchestration·

RAG Systems

Shipped

Enterprise knowledge assistant: vector embeddings, semantic search, hallucination tracking, and accuracy dashboards.

Vector Embeddings·Semantic Search·Python·

Prompt Engineering

Shipped

Role-based prompting frameworks that simulate senior PM reasoning — powering the AI PRD generator used across product cycles.

Role Prompting·OpenAI·Anthropic·

Automation Systems

Production

Content publishing and platform-operations pipelines for Listen2RE — 60% reduction in manual production effort.

Pipelines·n8n·CDN·

LLM Evaluations

Production

LLM-as-a-Judge quality gates: 87% pass rate, human flag rate driven from 31% to 13%. Evals as pre-launch infrastructure.

LLM-as-a-Judge·Quality Gates·Metrics·

AI Workflows

Shipped

Document analysis chains where each agent owns one reasoning step — automating work teams repeated hundreds of times weekly.

Document AI·Reasoning Chains·

Voice AI

Production

TTS pipeline producing audio-native learning sessions. The voice-quality upgrade outperformed every feature that quarter.

TTS·Audio UX·Pacing·

Product Experiments

Ongoing

A/B-tested engagement loops: push timing, streaks, session length. Habit mechanics measured, not guessed.

A/B Testing·Mixpanel·Retention·

Competency, instrumented

The same way I'd present product health: measured, visualized, and honest about levels.

sujit — product-execution — live
Product StrategyExpert92
RoadmappingExpert90
User ResearchAdvanced88
Data AnalysisAdvanced84
AI SystemsExpert90
Prompt EngineeringExpert93
Stakeholder ManagementAdvanced89
GrowthAdvanced85
ProductStrategyRoadmappingUserResearchDataAnalysisAISystemsPromptEngineeringStakeholderManagementGrowth

The knowledge library behind the products

Real working documents — PRDs, experiment docs, prioritization sheets. Click any to unlock. Happy to walk through any of them live in an interview.

Request a walkthrough

These are working documents from real products. I'll screen-share any of them and explain the reasoning.

Get in touch

What people who've worked with me say

Sujit owns the whole problem. He'd come back from campus visits with insights none of us saw in the data, turn them into a crisp PRD, and then actually sit with engineering until it shipped right.
EEngineering LeadWorked together at Zerton, 4 years

The full picture, one page deep

A decade of building, condensed for the 30-second scan and the 30-minute deep dive.

Resume — AI Product Manager

Experience, skills, projects, and achievements — formatted for hiring teams. Updated and ATS-friendly.

Download Resume

PDF · AI-PM-Sujit-Chankhore.pdf

Experience

  • CEO & AI Product Lead

    Zerton Education Technologies

    2023 — Present

  • Founder & Product Manager

    Zerton Engineering Services

    2015 — 2023

  • Trusted Photographer (B2B)

    Google Street View

    2016 — 2020

Skills

Product Strategy0→1 ExecutionLLM & Agentic SystemsRAG ArchitectureUser ResearchGrowth & GTM

Projects

  • Listen2RE — AI audio learning, 35K+ learners
  • B2B2C ERP — 26+ institutions, 42K+ users
  • Enterprise RAG knowledge assistant
  • AI-assisted PRD generator

Achievements

  • 77K+ users impacted across products
  • 60% content production effort cut with LLM pipeline
  • 8-person cross-functional team led
  • Published engineering research at L&T

Let's Build the Next AI Product Category

Open to Senior AI Product Manager roles, founding PM positions, and conversations about products worth building.