Skip to content
Nitesh Tiwari

Senior Product Manager

Growth × Consumer × AI

Retention is won before the retention metric.

I find where users drop before they reach value, fix the product before reaching for incentives, and set the quality bar before deciding where AI belongs — with an engineer's view of how it gets built.

~7 years product management · 10+ years technology · Edfora · Witzeal · Baazi Games

Open to Senior PM roles · Delhi NCR / Mumbai

Selected product outcomes

02Selected work

Four product stories, in depth.

All work

01  Strategic trade-offs

Doubt Resolution

Edfora · myPAT · Senior Product Manager · 2023–2026

Reduce resolution friction without scaling human support linearly.

  1. Signal

    ~70% of logged tickets were repetitive, pattern-based questions
  2. Selected

    A hybrid: guided hints, verified peer solutions, SME escalation
  3. Quality gate

    A 90% accuracy circuit-breaker on the hybrid; >90% accuracy maintained
  4. Piloted

    Two weeks, 10,000 JEE students, against a holdout; then a phased launch
  1. 02  Activation

    Onboarding Funnel Redesign

    Witzeal Technologies · Real-money gaming

    Only 12% of new users played a game on day one.

    Decision

    Five changes to the first 60 seconds, tested against a 30% control.

    Evidence

    12.2% to 25.4%

    Measured: Day-7 retention · control vs treatment, +13.2 pp · five-change bundle

    Bundle result · definition not recorded

  2. 03  Personalization

    Adaptive Practice Engine

    Edfora · Adaptive Practice

    Every learner was getting the same next question.

    Decision

    A 3PL IRT engine that matches each question to the learner's estimated ability.

    Evidence

    18% to 45%

    Observed: Assignment completion · Before/after · 2-year academic-cycle dataset

    Not attributed to the adaptive system alone

  3. 04  Behavioural systems

    Behavioural Loops & Gamification

    Edfora · Glorifire / Stakeholder

    DAU had been flat for two straight months, and teachers found early prototypes “too game-y”.

    Decision

    One configurable behavioural loop across students and faculty, rather than isolated point mechanics.

    Evidence

    DAU reported up 12–15% after launch; not causally attributed.

03Range

Same PM. Different domains.

Five consumer-growth capabilities, each proved in real-money gaming and in EdTech.

  1. Activation

    Gaming
    Five changes to the first 60 seconds
    EdTech
    A stuck learner's next step in minutes, not a day
  2. Retention

    Gaming
    D7 +13.2 pp for a five-change bundle, incl. ₹15 free games
    EdTech
    +18% relative uplift in D14 retention
  3. Experimentation

    Gaming
    30/70 rollout, ~50K users; a testing program
    EdTech
    2-week, 10,000-student pilot vs holdout
  4. Incentive economics

    Gaming
    Bonuses by segment ROI, retention held
    EdTech
    ~60% support-cost reduction (derived)
  5. Funnel diagnosis

    Gaming
    Loss traced to signup and OTP
    EdTech
    Completion read against learner ability

04Decisions

Smaller calls, same discipline.

Snapshots of product judgment, including the bets that were stopped.

All 11 decisions
  1. Signal
    Fewer than 6% of active match-day users used it.
    Decision
    Sunset the feature and move the effort to starting-XI notifications, injury alerts and head-to-head stats.
    Learning
    Users came for fantasy execution, not passive score consumption.
  2. Signal
    A fair number came back inconclusive or negative.
    Decision
    Use those results to change how later tests were scoped, instead of treating them as failures.
    Learning
    Experiments are not successful because they win. They are successful because they reduce uncertainty.
  3. Signal
    A system-generated myPlan is only useful if it is right, and faculty are the people who can tell.
    Decision
    Faculty confirm each myPlan as correct or incorrect, and accurate feedback earns 100 points, so verification is part of the behavioural loop rather than a chore outside it.
    Outcome
    Product-system evidence from the platform's screens, not an outcome. No accuracy or engagement result is attributed to it.

05AI

AI product judgment: problem first, model second.

The model is the last decision, not the first. A deterministic baseline, an evaluation, named failure modes, guardrails and a human override come before it, and sometimes the answer is not to ship it.

In the field. In doubt resolution, an AI auto-resolver was evaluated as a strategic option and not shipped as the first solution: academic trust and answer-quality risk outweighed its scale. A hybrid shipped behind a 90% accuracy gate. The decision behind it

Independent prototype · deterministic baseline

AI Learner Diagnostic

A deterministic, rules-only prototype; not an LLM. The baseline, the override and the evaluation harness are built. No evaluation has been run, and there are no real users or model results.

  • Implemented

    Baseline · Human override · Output schema · Evaluation harness

  • Designed

    Problem · Why AI · Evaluation cases · Architecture · Failure taxonomy · Launch gate

  • Planned

    Model decision · Context and retrieval · Monitoring · Model-based diagnosis

  • Needs input

    Educator label review · Latency and cost budgets · API access

Inside the AI Lab

06Career

Five roles. Each added a layer of the job.

  1. Built

    2014–2018

    Android Developer · Direct Create

    How software gets built.

  2. Shipped

    2019

    Program & Release Manager · PwC India

    How software ships.

  3. Measured

    2019–2022

    Product Manager · Baazi Games

    How users behave, and how to measure it.

  4. Grew

    2022–2023

    Product Manager · Witzeal Technologies

    How growth and monetization work.

  5. Personalized

    2023–2026

    Senior Product Manager · Edfora

    How a product adapts to each user.

Now

AI

Independent prototype · AI Learner Diagnostic (independent work, not a role)

The same discipline, applied to AI-native products.

The proof behind each phase

07About

I started by writing the product.

Before I owned roadmaps, I spent four-plus years as the only Android developer on a marketplace app, and then standardized releases for enterprise web applications. That order still shapes how I work: I scope with engineering, not around it.

About me, and how I work

“The drop wasn't motivation. It sat in signup and OTP verification, before the first game.”
Witzeal · diagnosis

08Contact

Hiring for a product role? Let's talk.

Open to Senior Product Manager and Product Manager roles, especially in growth, consumer products and AI. Delhi NCR, and Mumbai where it makes sense.