Dhruv Singhal

Product Manager & Builder — turning ambiguous domain problems into product decisions, specs, and shipped systems.

My story

Portrait of Dhruv Singhal

I pick up a domain by building something in it. That has been the pattern since my first internship. I get close enough to the problem to write the PRD, close enough to the data to build the eval set, and close enough to the code to build a first version when that is the fastest way to learn. Where someone built it with me, I say so. Then whatever breaks tells me what I got wrong in the spec.

At Omniful.ai, “find better prospects” turned into a scoring model built on firmographic and behavioural signals. Qualified prospects went from about 10 a day to over 200, and the work supported 10 client acquisitions (both self-reported; I have no artefact to share). At Read Riches I ran the founder's office side of content-led growth. I managed a 4-person research and content team and ran publishing experiments that contributed to a 4x retention improvement (self-reported). Different industries, same job: find the loop, instrument it, then turn the handle.

At Wipro TOPS I scoped an internal AI-powered enterprise workflow platform, along with workflows for an internal crew mobile micro-app and an internal records platform for non-crew staff, across 12+ aviation scenarios (self-reported). Most of that job was finding edge cases early enough that they became sprint tickets instead of incidents. It is also where I killed 37 low-signal dashboard charts and kept the 3 that actually drove a decision (self-reported). Most recently I was a product intern on the growth team at The Sleep Company (Jul–Oct 2026), on pre-launch work; the internal detail stays internal.

Outside of work there is Aarchid, which I co-built with Dilpreet Grover. It diagnoses plant health from a photo. I wrote the PRD and built the eval harness, and we shipped V1 together. Self-reported: the team ran an offline eval on a golden set (about 200 samples, as reported by the team); the result is withheld until the eval artefact or the co-builder's confirmation is available. No offline score could say whether someone would trust a diagnosis enough to act on it. Holding both of those at once is the part of the job I actually like.

What I am working on this month is on the now page. The books that shaped how I think are on the bookshelf, and the tools I use are listed on uses.

Education
B.Tech Electronics & Computer EngineeringJ.C. Bose University, 2022–2026

Four teams, four problems

The Sleep Company

Jul 2026 – Oct 2026

Product Intern, GrowthInternship

Growth intern on pre-launch product work for a D2C sleep brand. Internal systems and plans stay internal.

Wipro

Feb – Jul 2026

AI Product Intern, Wipro TOPSInternship

Scoped an internal AI-powered enterprise workflow platform, plus workflows for an internal crew mobile micro-app and an internal records platform for non-crew staff, across 12+ aviation scenarios (self-reported); surfaced edge cases early and moved specs into active sprint development.

Read Riches

2024 – 2025

Founder's OfficeConsulting

Ran content-led growth experiments, managed a 4-person research and content team, and contributed to a 4x retention improvement (self-reported) through better publishing cadence and feedback loops.

Omniful.ai

2024

Business Analyst InternInternship

Defined prospect scoring using firmographic and behavioral signals, scaled qualified prospects from about 10 per day to 200+ per day, and supported 10 client acquisitions (self-reported).

How I think

  • Outcome > Output

    I cut dashboards and features that do not change a decision, then align stakeholders around the few workflows that actually deserve priority.

  • Data Informs, Intuition Decides

    An offline eval score says how often the model is right on cases I chose. It cannot say whether people will trust the answer enough to act on it. Both questions matter.

  • Ship, Measure, Iterate

    Ship the smallest version that can be wrong in a useful way, then let what breaks rewrite the spec. Perfection is the enemy of learning.

Product, data, and enough engineering

Product

  • PRDs & Roadmaps
  • User Research
  • A/B Testing
  • Growth Strategy
  • Metrics Design

Data

  • Python
  • SQL
  • Pandas
  • Scikit-learn
  • Tableau

Technical

  • Next.js
  • React
  • TypeScript
  • Git
  • REST APIs

Soft Skills

  • Stakeholder Mgmt
  • Cross-functional
  • Presentation
  • Problem Solving

Highlights

  • Techstars Startup Weekend (DTU)

    3rd Place — Led problem discovery, market sizing, GTM and the pitch for a job-matching platform.

  • Smart India Hackathon

    Top 5.

  • Code Clash (VIT Vellore)

    Top 10.

GitHub activity

17Public Repos
1,557Contributions
1Day Streak
Top Languages
TypeScript (31%)JavaScript (23%)Python (23%)Elixir (8%)HTML (8%)

Last synced:

Operating principles

High-Ownership
Run my own release gates: the DeskTasks gate run of 15 Sep 2026 finished with zero failures
Close to the build
Wrote the Aarchid PRD and eval harness, and co-built v1 with Dilpreet Grover
Data-First
Every feature proposal comes with a success metric and a kill criterion
Fast Execution
Portfolio iterated from v1 (Jan 2026) to v6 (Sept 2026), each version sharpening positioning, content, or navigation; the changelog has every release.