Game 1 · 6 rounds · 60 seconds each
Spot the edge cases
A feature request lands and it sounds small. Find everything it breaks before the clock runs out, then see what you missed and why.
45 months building enterprise HR software · now at XLRI Jamshedpur
Spec it. Ship it. then find out what it broke.
I spent 45 months at Darwinbox working with engineering, QA and consultants on clear product specs and release reviews. Civil engineering at NIT Silchar before that; the PGDM in General Management at XLRI Jamshedpur now.
Case study 01 · Product analyticsLow adoption that pointed to a findability problemRead it →PlayTwo calls from my work
Make the calls I made, then see mine.
At Darwinbox almost every change raised two questions: what will this break, and will the people using it understand it. Each game puts you in front of six of those decisions, then shows the reasoning I would have used.
Game 2 · 6 rounds · about 3 minutes
Name the setting
A new HR admin has to find the right switch in their first week. Pick the label they would understand without asking anyone.
01Throughline
From site drawings to software specs.
-
2018 → 2022
Civil engineering
NIT Silchar · B.Tech · CGPA 8.99
Most of my family are civil engineers in government service, and I picked the branch over the others my JEE rank allowed. What I kept from it is the habit of settling a design on paper before work starts on site.
-
2022 → 2026
Product
Darwinbox · Product Analyst → Senior Product Analyst
I joined Darwinbox through campus placements. My role focused on defining product requirements and documenting them clearly for engineering, QA and consultants. Promoted in April 2025.
-
2026 → 2027
Management
XLRI Jamshedpur · PGDM, General Management
I wanted to build on my product experience with a broader business perspective. My manager and director, both XLRI alumni, encouraged me to apply. XAT 94.88 percentile, 99.67 in Quant and DI.
02Work
Darwinbox
Darwinbox sells HR software to large companies across Asia, the Middle East, Europe and North America. I worked on Core HRMS, the part every other module depends on: employee records, organisation structure, access control and bulk imports.
- Exemplary Contributor Award, 2024
- Promoted to Senior Product Analyst, Apr 2025
- Mentored 3 junior analysts
Specs
200+Product specs
Engineering-ready specs
The count is from Confluence: feature PRDs, change specs and edge-case documents I owned across Core, Travel and shared components. These included edge cases and acceptance criteria up front, giving engineering a shared implementation reference and QA explicit test conditions. I also worked on the Core HRMS roadmap with a Senior PM.
Analytics
700+features tagged
Usage tracking in Pendo
I defined a consistent, searchable tagging taxonomy and maintained it as the UI changed, keeping adoption records comparable. This supported three continuous years of records and gave roadmap discussions a usage-data reference alongside client feedback and escalations.
Research
24,000+survey responses
From survey to backlog
I set up in-app feedback for the new Profile page within days of its release. 24,000+ responses came back, and I sorted them with Engineering and QA into a ranked backlog for the same quarter. The page scored 4.3 out of 5.
Analytics
45%less time on key actions
Measuring the Dashboard revamp
I owned the measurement behind the Dashboard revamp, not its design. I instrumented the old dashboard, defined the key actions people came to do, timed them, and ran the same measurement on the redesign. Across 322,000+ users, time on those actions fell 45% while completions held or rose, so people were finishing faster, not giving up. The analysis became the business case for further platform investment.
Localisation
10+languages
Translation workflow
I owned the platform's translation workflow so the product team could manage routine updates independently. Each release I checked the languages I covered and used feedback from native-speaking clients to validate meaning and context.
Quality
100+issues caught
Pre-release reviews
I logged 100+ issues during pre-release reviews, helping resolve them before they reached clients. I also audited the settings screens and, with UX and engineering, regrouped 50+ settings to make their names and placement more consistent.
Inside a spec
Eight questions every PRD had to answer
The readers included founders, the Product Head, marketing, implementation consultants and release-note authors. I used their questions to clarify the document and make it a more useful reference for everyone involved.
Tick each one off · 0 of 8
03Case studies
Three problems, start to finish.
Case study 01 · Product analytics · Darwinbox
Low adoption that pointed to a findability problem
- The problem
- Pendo was available, but used mainly for surveys. Without consistent feature-level tracking, it was difficult to assess adoption after a release. Roadmap discussions relied heavily on client feedback and escalations.
- What I did
- I took Pendo on, defined a tagging taxonomy and tagged 700+ features, then built funnel and path views over the key Core flows: imports, lifecycle changes and settings.
- The call
- Low adoption could reflect limited demand or difficulty finding a feature. The path data pointed to the latter: people entered nearby flows but never reached those features. I prioritised plain-language setting names, help text and clearer placement before recommending replacement builds.
- What happened
- We shipped clearer names, help text and a reorganisation of existing settings to improve findability and comprehension. The same approach informed the broader settings audit described in Work.
Case study 02 · Localisation · Darwinbox
Making localisation work in context
- The problem
- Strings were translated individually and hard-coded. Feedback from native-speaking clients showed that some literal translations did not fit the context in which they appeared.
- The foundation
- An earlier centralisation effort brought translations into an internal page and established a shared foundation. Building on that work, I focused on a workflow for client-facing teams, with clearer navigation, defined access and ownership, and a consistent way to update strings across PHP and Angular.
- The call
- I mapped the recurring issues across settings and their dependencies and raised the workflow improvements for the roadmap. When the work was prioritised, I owned the PRD end to end. I built on the existing approach, focusing on clear access, ownership and update steps so the product team could manage routine translations independently.
- What happened
- Product teams could manage routine translation updates without engineering support each cycle. Turnaround improved from waiting for a sprint to the same release cycle, across 10+ languages.
Case study 03 · Strategy · XLRI Jamshedpur
An IPO-readiness deck where every number had to survive a question
- The problem
- Our team assessed for a panel whether Aster DM Healthcare's Gulf business was organisationally ready for an IPO. The argument rested on listed-peer comparisons, regulation and penalty figures, and much of that first turned up in secondary sources.
- What I did
- Before we presented, I traced every figure in our deck to a primary or named source, with one rule: anything that could not be traced was removed, not softened.
- The call
- One check involved two sourced figures: one listed peer's share-price decline since its IPO and another's twelve-month decline. The periods differed, so I found the second peer's offer price and restated both on the same basis. The review also aligned a penalty figure with the applicable company-size band and narrowed a legal claim to what the source supported.
- What happened
- Eight corrections completed before the panel, and a top 3 finish out of 20 teams. Where a figure could not be verified, such as the cost of a readiness programme, the deck identified who would need to assess it.
04Startup Lab
Five prototypes, each with a live demo.
Each one has a working demo you can click through, and the code for all five is public on GitHub.
01 · Settlement · Top 3, ReEnvision 5.0
Realium
A ₹18.6 lakh road bill, 60% paid the day after sign-off.
Public-works contractors wait months to be paid for finished work. Realium turns site-verified progress and an engineer's accountable sign-off into a claim a bank will fund. In the worked example 97% of the bill reaches the contractor, against 92.7% if they borrow informally at 18%. Our team placed 3rd of 15 at ReEnvision 5.0, XLRI's Digital Transformation Conclave, in July 2026. My slides covered the three layers below.
- L1EvidenceWork verified on site
- L2AuthorityAn accountable sign-off
- L3LiquidityA bankable instrument
-
02 · Memory
Heirloom
Keeps a firm's knowledge as records the firm owns, so it survives people leaving and AI vendors changing. The demo vault lets you ask as three different roles, make a partner leave, and export everything as one file.
Still unproven: whether anyone buys memory on its own rather than bundled with an assistant. The likeliest first buyer is a regulated firm that has lived through a costly migration.
-
03 · Commerce
Turnstile
Tells people, AI shopping agents and scrapers apart in a store's traffic, then serves each the right thing: the normal page, a clean product feed, or nothing. The demo lets you tune the rules and watch six sessions re-sort.
Still unproven: whether shoppers want an assistant buying for them at all. My bet is that clear controls and a visible receipt would earn that trust.
-
04 · Simulation
Windtunnel
Plays a price increase out 500 times across three customer groups and shows who leaves and how often the move backfires. A flat 12% rise holds up; at 30% it loses money in more than half the runs.
Still unproven: the customer behaviour, which is set by hand. The next step is checking the model against real past price changes.
-
05 · Agents · Manakriti 3.0
Heir
Student committees change every year, and what the old team promised sponsors, speakers and campus offices leaves with them. Heir is five agents on one ledger: they read the club's own mail and files, catch what the handover doc missed, plan the next event and draft every email for a member to approve. The demo is a game: take over a club, then run it with Heir. Built with Team Ghost Protocol for XLRI's Manakriti 3.0 agent challenge.
Still unproven: it has only run on a mock club. Calls and DMs reach it only if members forward them, so a pilot has to show they will.
05Record
Recognition and life outside work.
Recognition
- 20263rd of 15 teams, ReEnvision 5.0XLRI Digital Transformation Conclave, with Realium
- 2026Top 3 of 20 teamsAster DM Healthcare, IPO readiness assessment
- 2024Exemplary Contributor AwardDarwinbox, for cross-team delivery
- 20192nd, AutoCAD eventNIRMAAN '19, NIT Silchar
- 20183rd, ARCHENOVA smart city moduleTecnoesis '18
- 20182nd, Model MakingSwachhata Hi Seva 2018, Ministry of Human Resource Development
Campus
- 2026Prodzion, XLRI's product clubElected club member, 8 of 118 students
- 2026AURORA 4.0 organising teamEnd-to-end panel, travel and hospitality for 20+ industry leaders
Off the clock
- KeysKeyboardistPlayed with the cultural team at inter-college fests, including IIT Kharagpur and Guwahati
- 100+Self-taught musicianSongs uploaded to my YouTube and Instagram accounts
06Ask me
Four questions I get asked.
CareerWhy move from civil engineering to software?
Civil engineering is still close to me. Most of my family works in it, and I still enjoy seeing how things get built. Darwinbox came to campus early, and I applied out of curiosity. Moving into product work showed me how the habit of planning a build in detail could carry over to software. A PRD does for a software team what a good drawing does for a site. I found a new setting in which to use that skill.
CareerWhy an MBA after 45 months at Darwinbox?
I wanted to build on my product experience and develop a broader business perspective. Working with my manager and director helped me understand how decisions bring client needs, team priorities and business goals together. Both are XLRI alumni and encouraged and supported my application.
Working styleHow do you keep engineers unblocked?
I documented edge cases, acceptance criteria and dependencies in each PRD. Before the coming sprint, I coordinated with the developers taking up the work and scheduled a knowledge transfer. The document and handover established a shared understanding, while follow-up discussions focused on decisions that needed clarification.
JudgementWhat do you do when your own feature does not land?
Features I owned reached anywhere from 5% to 60% account-level adoption, and one I had high hopes for was at the low end. I initially thought rollout might explain it, then reviewed the drop-off data and reconsidered the design. The evidence did not support the value I had expected it to deliver. I would rather test my assumptions during planning than have clients encounter avoidable limitations after release.