Ridhi Jolly

Shipped: KaDeep Studios, TestStudios, Accessibility, Skills-first, Plan, then drive, WCAG 2.2, IEEE published, India · Canada · UAE.

Work

Three products. One bet: the model is not the product.

Decision, tradeoff, miss. Open a study when you have a minute.

Proof

The product bet

  1. Playbooks first

    Deterministic skills take the common path.

  2. True miss?

    If the UI actually moved, then spend the model.

  3. LLM second

    Cost sits on novelty — not on everyday churn.

  • 100+

    QA + product seats

    Analytics adopted internally

  • 25%

    Faster release

    Three core modules

  • 89

    xactload.com score

    0 Critical / 0 Serious

  • IEEE

    Peer-reviewed

    94.6% accuracy · 97.1% F1

Experience

How I operated.

Engineer first. Then PM. Same company, same products — a different seat in the room.

KaDeep AI

Full-time · 9 mos

Mumbai, Maharashtra, India

  1. AI Product Manager

    Aug 2026 – Present · 2 mos

    Hybrid

    • Software Product Management
    • Artificial Intelligence (AI)
    • Own the product loop across Studios, TestStudios, and Accessibility — one roadmap with founder and engineering, not four parallel plans.
    • Skills-first, LLM-second: playbooks recover first so model cost sits on true misses. Analytics adopted by 100+ QA and product seats; 25% faster release across three modules.
  2. Product Engineer

    Jan 2026 – Aug 2026 · 8 mos

    On-site

    • Product Management
    • Led end-to-end ownership of the company-wide accessibility module — defined implementation strategy, standards, and workflows, and drove cross-team execution from planning to production.
  • 4

    Surfaces, one roadmap

    Automation, healing, Accessibility, TestStudios — sequenced, not parallel.

  • 100+

    Seats on analytics

    Pass-rate, heal, release-readiness. 80+ bugs caught before they blocked a ship.

  • 9

    Months at KaDeep

    Engineer first, then PM. Same products. Different seat in the room.

How I work

Product judgment, then the model.

I sit with engineering, QA, and the person who has to defend the release. Frame the problem, put a cost on the model, ship a loop a team will actually run.

The decisions live in the case studies. I want the next seat in AI, platform, or developer tools — India, Canada, or the UAE.

Ridhi Jolly

Frame first

ICP before the feature list.

Cost the model

Playbooks run. The LLM pays for true misses.

Language is a feature

Evidence from this scan. Never a fake stamp.

Own the miss

If I was wrong, it is on the page.

Can I sit in this room?

Product

  • Problem framing
  • Roadmapping
  • PRDs
  • RICE
  • Packaging & GTM

AI product

  • Skills-before-LLM
  • Cost per miss
  • NL authoring
  • Agent loops
  • Evals

Compliance

  • WCAG 2.2
  • ADA / 508
  • EN 301 549
  • VPAT

In the repo

  • TypeScript
  • Next.js
  • Playwright
  • Figma
  • Jira

IEEE Published · Mar–May 2025

Multi-Model Sentiment Analysis for E-commerce

An ML pipeline that turns reviews into merchandising insight. Peer-reviewed. Accuracy is on that evaluation set — not a production SLA.

Read the IEEE paper (opens in a new tab)

94.6%

Accuracy

97.1%

F1 score

Education

Bennett University.

B.Tech CSE. The rest is on the resume.

Sep 2022 – Jun 2026

Bennett University

B.Tech, Computer Science and Engineering

  • Dean’s List
  • CGPA 8.32 / 10
  • Dean’s List

    Bennett University — selective academic honour.

  • HackEye’24 runner-up

    National hackathon. Shipped and pitched in 24 hours.

  • LeetCode 1679+

    500+ problems. AIR 282, Coding Ninjas Premier League DSA 2023.

Before KaDeep

Two 0-to-1 cuts.

Contact

Let’s talk about the role

Hiring an AI or platform PM who has already shipped generation, healing, a no-code agent, and accessibility — write. India, Canada, UAE.

Send a note. It lands in Ridhi’s inbox. Nothing is stored on this site.