AI ProductLive Build · Manufacturing

ShopFloor AI — AI-powered production tracking for manufacturers

Identified the gap in manufacturing visibility tools — shop floor operators lack real-time production tracking, inventory awareness, and a natural-language way to query floor status. Designed and shipped an AI-powered MES with voice, Kanban, and live inventory tracking.

• Built a Kanban production board with drag-and-drop work orders across Queued, In-Progress, QC, and Done stages — giving plant operators a single real-time view of the shop floor.
• Designed a bilingual AI assistant (Hindi + English) for natural-language floor queries — 'Aaj kitne orders pending hain?' — making the tool accessible to operators regardless of language preference.
• Shipped real-time inventory tracking with auto-deduction on order completion and live reorder alerts, reducing the risk of production stoppages from stockouts.
• Included a live demo mode (no sign-up required) with real manufacturing data, lowering the barrier to evaluation for potential users.

Live · AI-Powered MES · Voice in Hindi & English · Zero sign-up demo

Open live app ↗

AI ProductCase Study · PRD

ChurnGuard AI — AI-powered churn prediction and proactive retention

Identified why SaaS companies lose revenue to churn they could see coming — lagging indicators, reactive support, and no automated save workflows. Designed an AI system that scores customer health in real time and triggers proactive retention plays before cancellation intent becomes a cancellation event.

• Defined a customer health scoring model combining behavioral analytics (product usage, login frequency, feature adoption) with CRM signals to surface at-risk accounts before they churn.
• Designed agentic retention workflows using n8n — triggered automatically when a customer's health score drops below threshold, routing to the right save play (outreach, discount, success call) without manual intervention.
• Specified proactive retention metrics: save rate, churn rate reduction, time-to-intervention, and health score accuracy — tied directly to ARR impact.
• Scoped the product from signal detection through workflow orchestration to outcome tracking, covering the full retention loop in one system.

AI-driven · Churn Prediction · Proactive Retention · Agentic Workflows · n8n

Open case study Open PRD

AI ProductPRD · V1.0

PM DocForge — AI-powered PM document generator

Designed and shipped an AI web app that transforms a brief product description into a full suite of PM documents — PRD, Design Doc, Technical Spec, GTM Brief, User Stories, and Release Notes — in under 60 seconds, no signup required.

• Identified the core problem: writing a single PRD takes 2–6 hours; existing tools require manual editing or enterprise subscriptions.
• Defined a frictionless two-step flow (Describe → Choose Docs) powered by Claude AI with instant DOCX/PDF exports.
• Set 90-day success metrics: >65% activation rate, 500+ docs/week, <90s time-to-download, NPS > 45.
• Scoped V1 in/out boundaries — no accounts, no collaboration, no integrations — to ship fast and validate core value.

Target: 500+ docs generated per week with NPS > 45 by end of Phase 3

Open PRD case study

AI ProductPRD · Strategy

PolicyPilot — AI-first health insurance marketplace

Diagnosed why AI insurance assistants like PolicyBazaar's PB Buddy stall before purchase, then specified a smoother product where every recommendation is actionable and the journey from first question to active policy completes in one flow.

• Identified three root causes of drop-off: dead-end plan recommendations (plain text, no buttons), fragmented intake, and no chat-to-checkout handoff.
• Designed a 5-stage continuous flow — Discover → Advise → Decide → Buy → Serve — where context carries forward and nothing is re-entered.
• Specified an actionable-recommendation contract: every plan rendered as a card with View / Buy buttons, making dead-end recommendations structurally impossible.
• Defined success metrics: >40% recommendation-to-action rate, >60% flow completion, 2–3× chat-to-purchase conversion in under 10 minutes.

Target: 2–3× baseline chat-to-purchase conversion · >55% assistant containment

Open PRD case study

AI ProductCase Study · Live Build

askDr.ai — RAG-powered health information assistant

Identified why generic AI chatbots fail at health queries — hallucinations, no citations, no safety guardrails — then designed and shipped a grounded health assistant using Retrieval-Augmented Generation, verified medical sources, and a safety-first architecture. Built solo in 4 days at $0 on free tiers.

• Chose RAG over a plain LLM to make every answer source-traceable: responses cite the specific openFDA drug label or MedlinePlus guideline they came from, making hallucination structurally constrained.
• Designed safety-first architecture: emergency queries (chest pain, self-harm, overdose) bypass RAG entirely and return crisis guidance before any AI processing begins.
• Instrumented a full Mixpanel funnel (Page View → Category → Message → Response); 28.57% drop-off at landing drove the decision to add suggested starter questions and a clearer CTA.
• Built and ran a 12-test automated eval suite — initial 7/12 score exposed missing safety patterns; fixed before launch and shipped at 12/12 across retrieval quality, grounding, safety, and hallucination resistance.

12/12 eval suite · Safety-first RAG · Shipped in 4 days · $0 infra cost

Open case study Live app ↗

AI ProductPRD · v1.0

askDr.ai — Product Requirements Document

Full PRD for a RAG-powered health information assistant — from problem statement and user personas through P0/P1 requirements, safety & compliance rules, success metrics, data architecture, 4-phase release plan, and risk register.

• Defined three user pain points (hallucinating AI, information overload, no safety net) and two primary personas — the Informed Patient and the Caregiver — with explicit anti-personas to keep scope tight.
• Specified P0 requirements enforcing RAG grounding, a pre-RAG emergency screen, no-diagnosis policy, and mandatory citations on every response as non-negotiable trust constraints.
• Set five success metrics with targets: grounding rate >70%, safety pass rate 100%, engagement >3 messages/session, funnel conversion >30%, eval score 12/12.
• Documented a 4-phase release plan (Medicine RAG → All Categories → Vision + Streaming → Analytics + Eval) with clear ship criteria per phase and a risk register covering critical, high, and medium risks.

Shipped · All P0 requirements delivered · 12/12 eval · Live at ask-dr-ai.vercel.app

Open PRD

StrategySWOT deck

WhatsApp SWOT analysis and Smart Priority Inbox strategy

Analyzed WhatsApp's strengths, weaknesses, opportunities, and threats, then proposed AI-first usability improvements to reduce message overload.

• Identified message overload, limited search, and chaotic groups as core usability gaps.
• Recommended AI chat summarization, smart priority inbox, and context-based notifications.
• Defined metrics like meaningful conversations, time to find, response rate, and notification fatigue.

Expected outcome: 20% drop in notification fatigue score over one quarter

Open PDF case study

TeardownProduct analysis

Cvent product teardown and enterprise event SaaS analysis

Deep-dived into Cvent's event management platform, pricing model, user personas, activation friction, AI strategy, and enterprise event workflows.

• Mapped the end-to-end event manager journey across onboarding, registration, go-live, onsite, and analytics.
• Identified friction in 3-step publishing, module complexity, email tooling, pricing opacity, and India-market gaps.
• Prioritized fixes like Smart Publish, event-type onboarding, transparent pricing, and CventIQ opportunities.

Target: Cut time-to-first-event under 30 min for new enterprise users

Open PDF teardown

FemtechOnboarding

NUA onboarding case study and subscription conversion strategy

Studied NUA's app onboarding, period tracker activation, subscription discovery, loyalty loop, and marketplace-to-app retention gap.

• Found friction in the 10+ question plan quiz, empty tracker state, spend-only loyalty, and buried subscription CTA.
• Mapped impact to MRR, subscription conversion, tracker activation, Comfort Club progression, and pause/cancel rates.
• Recommended surfacing subscription earlier, progressive onboarding, engagement-based rewards, and symptom-led discovery.

Expected outcome: +12% subscription conversion within 2 sprints

Open PDF case study

Live BuildJob Tracker · Web App

Job Application Tracker — stay on top of every opportunity

A focused web app to track job applications end-to-end — add roles, log status updates, set follow-up reminders, and see your pipeline at a glance.

• Built to solve a real personal pain point: losing track of applications across dozens of companies.
• Clean dashboard showing pipeline stage (Applied → Interview → Offer → Rejected) per application.
• Live and free to use — no setup, just sign in and start tracking.

Live product · Try it at job-tracker-nu-five.vercel.app

Open live app ↗