Associate Product Manager · AI Products · Product Strategy

Building better products starts with asking better questions. 

I'm Sivamanikandan R. I solve user problems by validating ideas with real customers and turning what I learn into AI products people keep using.

See the work Let's talk →
questions extracted
80K+questions extracted
AI Extractor customers
50+AI Extractor customers
Review Mode customers
70+Review Mode customers
Bluebook Insights users
500+Bluebook Insights users
UnderstandDefineValidateIterate

01 / 04

What problem, for whom, and how do we know?

01About · Product approach

A little about me

I'm a product professional who enjoys figuring out why something isn't working before jumping into how to fix it.

My background combines an MBA in Business Analytics & Supply Chain Management with hands-on work in operations and analytics. At TeachEdison / EdisonOS I owned three AI and education products, working directly with customers, designers and engineers.

How I approach product work

  1. 01Ask why before what

    Review Mode began as an AI tutor feature. Asking why it was needed led to the real problem: students need to see where they went wrong.

    Review Mode

  2. 02Validate before scaling

    I didn't assume the AI Extractor beta was right. 5–7 customers used it before the first proper release.

    AI Extractor

  3. 03Watch what people actually do

    Users extracted whole PDFs but used about 80% of the output, which led to the page selector. Waiting complaints led to completion notifications.

    AI Extractor

02Selected work

Three products, three stages

A flagship AI product I grew, a 0 → 1 product I helped found, and a mature product I looked after.

Flagship01

AI Extractor

Helping tutors bring their own content into the platform in minutes instead of the better part of an hour.

45 → 3–5 minper question

Read the story

Founding product work02

Review Mode

Showing students and tutors exactly where a student went wrong, so nobody loses context between 1-on-1 classes.

0 → 1founding product

Read the story

Mature product03

Bluebook Insights

A free community product that turns College Board Bluebook exam results into detailed reports.

500+users

Read the story

Case study 01 · Flagship · AI product · Growth & iteration

AI Extractor

Helping tutors bring their own content into the platform in minutes instead of the better part of an hour.

My role
Product owner: discovery, requirements, customer validation, prioritisation and release coordination
Built with
A prompt-engineered backend using Google Gemini, implemented by our engineering team
Paper trail
7 PRDs

North Star

Reduce the time and effort required for tutors to bring their own content into the platform.

01Problem

Getting content in was the bottleneck

Tutors bringing their own content into the platform had to get every question into the live product manually, which took roughly 45 minutes per question.

02Baseline

45 minutes down to 3–5

per question, manual~45 min
per question, with AI Extractor~3–5 min

≈ 89–93% less time per question

03Discovery → redesign

From a rough beta to a first real release

  1. Discovery

    The team began by collecting many different PDF structures and talking with customers about the materials they had and what they needed simplified. Customer calls, support tickets, newsletters and direct conversations surfaced the recurring problems.

  2. Beta

    I joined when the first version could already output extracted files in a suitable format, but had no polished interface. I didn't assume that first version was right.

  3. Customer validation

    I had 5–7 customers use the beta, collected their feedback, identified the problems and made changes.

  4. First release

    Those changes went into the first proper release.

  5. UX redesign

    The team expanded with a designer, and we redesigned the UI/UX to simplify the extraction workflow.

04Product decisions

What we noticed, and what changed

  1. Exam expansionStarted with one exam type and expanded to multiple exam types.
  2. Page selectorUsers were extracting whole PDFs even though only around 80% of the extracted content was ultimately useful. I proposed page selection to cut unnecessary processing and token consumption.
  3. Exam-type UXUsers were confused by the exam-type choices, so we simplified that part of the UI.
  4. Completion notificationsUsers reported long waits. The deeper problem was that they believed they had to stay on the loading page, so we added notifications to let them leave and come back when processing finished.
  5. Background queueingAdded background processing for multiple folders, one after another.
  6. Intelligence ExtractAdded AI answers/solutions and tagging of questions to the relevant skills and exam domains.
  7. Landing pageVibe-coded the product landing page at extract.edisonos.com.
  8. Customer loopSent biweekly newsletters, joined customer calls and reviewed support tickets continuously.

05Scale

From a handful of customers to 50+

When I took over

customers
~5customers
questions extracted
4–5Kquestions extracted

Now

customers
50+customers
questions extracted
80K+questions extracted
PDFs processed
700+PDFs processed

06Quality

Accurate enough to come back to

All observed current customers are repeat users: they keep updating content for new exams and cohorts.

extraction accuracy
98%extraction accuracy
customer satisfaction
4.9/5customer satisfaction

07Learnings

What I took away

A useful AI feature is not just about model capability. The surrounding workflow, user confidence, error handling, history, notifications, and export options can determine whether users actually adopt it.

Next case studyReview Mode

Case study 02 · Founding product work · 0 → 1

Review Mode

Showing students and tutors exactly where a student went wrong, so nobody loses context between 1-on-1 classes.

Stage
Founding product work, from 0 to 1
Design
5 UX iterations · high-level UI/UX in ~30 days
Customers
Involved throughout discovery and iteration

North Star

Understand where the student is actually going wrong, especially when gaps between 1-on-1 classes cause students and tutors to lose context.

01Starting question → why

Asking why until the real problem showed up

The work initially started around an AI tutor feature. By repeatedly asking why the AI tutor was needed, we found the deeper problem.

Started withAn AI tutor feature
Why?
Why?
Why?
Real problemStudents need to understand where they went wrong on each question, not simply get an answer

02Opportunity

An underserved space

That revealed an underserved, largely untouched product domain where competitors were not doing the same thing.

03Student problem → tutor value

Two users, one product

The B2B SaaS model created a unique challenge. Tutors were the primary business users, but the feature needed to create direct value for students in a way that improved the tutor's workflow.

04Experience design

What a student and tutor see

  • Confidence score
  • Time and reason for each missed answer
  • Skill accuracy tags
  • Information shared with both student and tutor
  • AI tutor support
  • Whiteboard notes visible to the tutor

5 UX iterations, with the high-level UI/UX completed in about 30 days.

05Customer validation

Built with customers in the room

Customers were involved throughout discovery and iteration, and we iterated with them before launch.

06Launch · current usage

The first version is live

customers
70+customers
students
200+students
typical engagement
Weeklytypical engagement
questions reviewed per session
~7questions reviewed per session

These are current adoption numbers for the first version, not post-launch improvement claims. Session length varies because the experience is student-driven.

Next case studyBluebook Insights

Case study 03 · Mature product · Ownership

Bluebook Insights

A free community product that turns College Board Bluebook exam results into detailed reports.

Stage
Mature product
My role
Owned product direction, features and release coordination
Paper trail
3 PRDs

01Mature product

A free product the community already uses

Bluebook Insights provides detailed College Board Bluebook exam result reports for the community, free.

total users
500+total users

02Ownership

Looking after a product that already works

  1. Customer adoptionActively promoted the product to customers.
  2. Feedback & bugsIdentified bugs and coordinated fixes at appropriate intervals.
  3. PrioritisationOwned product direction, features and release coordination.
  4. Bluebook 11Bluebook 11 support was added while I was looking after the product.
  5. UX improvementsMade UI changes to create a more seamless user flow.
  6. Ongoing ownership3 PRDs associated with the product work.

03Learnings

What I took away

Good product decisions often come from removing confusion and improving the path to value, not from adding more features.

How I work · across all three products

The customer work behind all three products

products owned
3products owned
customer calls
70+customer calls
customer interviews / findings
20customer interviews / findings
customers contacted
100customers contacted
support tickets personally owned
100+support tickets personally owned
  • 5-person product team: 1 product, 2 designers, 2 engineers
  • 200+ support tickets analysed overall
  • Coordinated every release across the three products
  • 7 PRDs for AI Extractor · 3 for Bluebook Insights
  • 5 UX iterations for Review Mode
  • 2 beta experiments run

03Experience

Where the work happened

Associate Product Manager · TeachEdison / EdisonOS

Sep 2025 – Aug 2026

Owned three education products: AI Extractor, Review Mode and Bluebook Insights. I worked from customer discovery and requirements through UX collaboration, testing, launch and iteration, inside a 5-person product team of 1 product person, 2 designers and 2 engineers.

Before product

  1. Sep 2024 – Feb 2025

    Chartwells Higher Ed

    Inventory Manager & Supervisor

    Managed day-to-day operations for a university food-service environment, including inventory planning, ordering, team coordination, scheduling, and process improvement across a 30-person team.

    Reduced food waste by 90%, improved order accuracy to 100%, reduced processing time by 25%, increased shift coverage by 30%, and reduced overtime costs by 20%.

  2. Jun 2023 – Aug 2023

    Tata Motors

    Sales & Data Analyst Intern

    Supported sales and customer-focused initiatives using data to identify trends, improve reporting, and support promotional activities.

    Contributed to database efficiency, customer engagement, and data-informed promotional initiatives.

  3. Graduate

    Northern Kentucky University

    MBA – Business Analytics & Supply Chain Management

    Combined business analytics, supply chain, strategy, and management coursework with practical problem-solving and data-driven decision making.

    GPA: 3.66 / 4.00

04Product thinking

How I actually work through a problem

Four questions I keep coming back to, each with a moment from the work above.

Step 01 · Understand

Who is the user, what problem are they having, and what evidence do we have?

In practice

Collected PDF structures and talked to customers about the materials they actually had.

From AI Extractor

05Contact

Let's build something useful.

I'm open to Product Manager, Associate Product Manager, and product-focused opportunities where I can solve meaningful user problems and help turn ideas into useful products.

Get in touch