Aug 2025 – presentAI
Aarchid
Photograph a sick plant, get a diagnosis you can act on.
Live92%Diagnosis Accuracy (200-sample eval)
Dhruv Singhal · Product Manager & Builder
I scope, spec, and ship the V1 myself.
Currently: Product Manager Intern, Growth · The Sleep Company
E-commerce and AI products — stakeholder input, user research, and product specs turned into decisions teams can build.
Five products, built end to end
May 2026 – presentTechnical
KiteEdge
The analytics desk a fund gets, for a retail Zerodha trader.
Internal43+Technical Indicators
2026 – presentProduct
ExperimentHub
Self-hosted A/B testing with deterministic variant assignment.
Open source50+Tenant-Scoped API Routes
2026 – presentProduct
DeskTasks
Your task list, pinned behind every window on the desktop.
Live540+Test Assertions (unit + e2e)
Mar 2026 – May 2026Technical
TCS NQT Prep Hub
Free, offline-capable TCS NQT prep with 424+ practice questions.
Live424+Practice Questions
Also built
- 300+Organic Users
Hackmate — Co-Founder Matchmaking Platform
Helps solo founders find the co-founder they'd actually build with. Swipe-based matching shaped around real founder-discovery pain points, and iterated with the community it grew in.
- 575Live RLS Policy Checks
Better-Half — Cycle-Aware Companion
Helps long-distance couples stay in sync when they can't read the room in person. Cycle phase, mood, and stress drive daily guidance for her, him, and us. Privacy is enforced at the database rather than the UI: 575 live RLS checks, offline-first, $0/month infra.
- ~15%Churn Reduction
Customer Churn Analysis
Told the retention team which customers were about to leave, early enough to do something about it. A predictive model that surfaced at-risk users and cut monthly churn by ~15%.
- 95+Lighthouse Score
This Portfolio — Meta Case Study
Treated my own portfolio as a product: spec and success metrics before any code, then iterations driven by real feedback. Ships at 95+ Lighthouse, with the whole decision trail written up as its own case study.
Measured outcomes
92%Diagnosis Accuracy (200-sample eval)
43+Technical Indicators
9Case Studies
14Articles Published
How I think
- AI · 8 min read
Shipping LLM Products Starts With the Eval Harness, Not the Prompt
A prompt is an artefact. An eval harness is a product. Here's how I scope LLM features so the output doesn't surprise users — or me.
- Career · 6 min read
Why PMs Should Code (And When to Stop)
I have been the PM who could not see why a “simple” feature took three sprints, and the PM who accidentally rewrote an engineer’s PR. Where the useful middle sits.
References
“Dhruv brought structure to messy aviation workflows quickly. He turned SME inputs into clear product specs, surfaced edge cases before handoff, and consistently made cross-functional conversations easier for engineering and QA.”
12+Workflow Scenarios Scoped
Where to go next
- AI
The PRD Is Dead, Long Live the Eval Set
I wrote a PRD that passed review with no blocking comments, and then nobody could tell me whether the feature worked. Prose cannot spec a probabilistic system.
- AI
Choosing a Product Without AI in It
I ranked my next side project on five criteria. The AI-heavy ideas scored highest on excitement and lowest on everything that predicts a finished tool.
- Product
From 'It Looks Off' to a Ranked Tracker: Running a Figma-to-Live Parity Audit
A design review ended with it looks off and nothing anyone could pick up. How I turned that into 45 ranked, located, owned rows.
- Data
Cohorts Before Dashboards
Two products can share a DAU line and have opposite futures. A retention curve cut by acquisition cohort tells you which one is actually compounding.
