Prabhav Mhapne
Problem-first full-stack product manager with IT/OT background and strategic fundamentals. I translate business goals & customer needs into functional user stories, features & products that solve real-world problems.
Skills & Expertise
Product skills from discovery to delivery, powered by hands-on AI development
Career Journey
From engineering foundations to AI product management
Blumenbecker
Automation Engineer
2017 – 2021
Vanderlande
Controls Proj. Engineer
2021 – 2022
RWTH Aachen University
MSc. Tech. Innovation Mgmt.
2022 – 2024
Porsche Digital
PM Intern + Werkstudent
2023 – 2024
Digitale Leute School
PM Bootcamp
2025 – 2026
AI PM Capstone Project
Verifyr & RAG Systems
2025 – Present
Featured Products
Functional prototypes demonstrating end-to-end product ownership, AI integration, and measurable impact
AI Shopping Advisor Platform
Conversational AI advisor helping fitness & health enthusiasts make confident purchase decisions through curation of product data, manuals, reviews, test results, that answers all your pre- & post-purchase questions and provides neutral, source-backed and personalized recommendations.
E-Commerce QA Bot
Functional RAG-based Knowledge chatbot delivering accurate answers from 250+ pages of product documentation, setup manuals & curated tests/reviews in seconds.
GenAI Feature(s) in CarPlay
Beta Program validating GenAI features before production rollout. Dual goals: validate product quality through real customer feedback and create communication potential through innovation.
Featured on LinkedIn
Insights on product strategy, AI implementation, and lessons learned
Prototyping for Validation: Building Verifyr Health-Tech Product
Shared learnings on using prototypes to test assumptions rather than just create mockups. Focused on user research, hypothesis-driven testing, and intentional MVP scoping for a health-tech purchase advisor.
Read on LinkedIn →Key Insights from Digitale Leute Summit 2025
Spent a day immersed in AI and product talks in Köln. From GenAI fundamentals and AI muscle development to cutting customer feedback analysis from weeks to hours—here's what stuck with me from industry leaders at Google, Kameleoon, Taxfix, Scout24, SIXT, and Make...
Read on LinkedIn →Master's Thesis: GenAI in Agile Product Teams
Wrapped up my MSc exploring the potential of integrating GenAI in agile product working teams at RWTH Aachen University. Key takeaways: AI boosts human capabilities in productivity, innovation, and decision-making through augmentation, not replacement...
Read on LinkedIn →Porsche 6-Hours-Run for Ferry-Porsche-Stiftung
Participated in the Porsche 6-Hours-Run alongside an amazing team at Porsche Digital, supporting children and teenagers. Completed 11 rounds of 911m laps. Grateful for the experience and proud of what we achieved together...
Read on LinkedIn →View All LinkedIn Posts
Follow me on LinkedIn for more insights on AI product management, discovery practices, and lessons learned from building products...
Visit LinkedIn Profile →Let's Connect
Open to PM roles & opportunities, or just a conversation about product and AI.
AI Shopping Advisor Platform (Verifyr)
Problem
Wearables shoppers experience decision paralysis comparing tech products across 20+ browser tabs with complex jargon, contradictory reviews, and commission-biased recommendations. Health tracking is personal—users don't trust generic advice.
Solution: Questions → Recommendations & Comparison → Decision
Conversational AI platform that asks smart questions, compares 3-4 devices side-by-side, and provides neutral recommendations based on curated product manuals, reviews, and test results. Includes Q&A chat for pre- and post-purchase support. 100% commission-free.
Execution & Impact
Discovery: 30+ user interviews identified 67% experience decision paralysis. Validated €450K+ revenue potential through market analysis and willingness-to-pay research (70% would pay €5-10/month).
Design & Validation: Built 50+ Figma screens, ran 4 usability test rounds. Launched landing page achieving 8% conversion (16/200 visitors)—above 2-5% industry benchmark. Developed RAG chatbot prototype (3-7s response time, 85% relevance accuracy).
Development & Deploy: Built production RAG backend (FastAPI, multi-model LLM with Claude/GPT-4o). Deployed bilingual landing page (Netlify, GDPR-compliant). Overcame technical challenges (CORS errors, language switching bugs) through independent debugging.
Key Learnings
Validate before building: Landing page + interviews saved 6+ months of building the wrong product. Niche down: Starting with wearables (not all products) made value prop clearer—8% conversion proves focus works. Test early: Major UX pivot discovered in round 3 of usability testing—structured form → conversational flow increased satisfaction 60%.
Interactive Prototype
Explore the conversational AI flow designed in Figma:
E-Commerce QA Bot
Problem
Online shoppers waste hours jumping between product pages, 20+ tabs, review sites, and forums. They need unbiased, instant answers from actual product documentation—all in one place.
Solution: RAG Knowledge Bot
Product-specific knowledge bot that retrieves answers from specs, user manuals, setup guides, and verified test results. Provides pre-purchase support (product comparison, spec translation) and post-purchase support (setup, troubleshooting)—transforming documentation into on-demand assistance.
Execution & Impact
RAG Pipeline: Built FastAPI backend with multi-model LLM (Claude, GPT-4o) and Langfuse observability. Indexed 250+ pages across 2 pilot products.
Evaluation System: Created 40-case golden dataset (CRISP framework: factual, comparison, complex, edge cases). Implemented LLM-as-a-Judge with custom domain scores. Result: 87% faithfulness, 85% answer relevance (RAGAS metrics).
Performance: 3-7s response time with hybrid search (semantic + keyword). Cost-optimized with prompt caching (~€0.05/query).
Key Learnings
Eval-first approach works: Building test cases before scaling prevented quality issues. Multi-model strategy: Claude for complex reasoning, GPT-4o for speed—saved 30% on costs. Hybrid search matters: Semantic alone missed exact model numbers; keyword search fixed this.
GenAI Feature(s) in CarPlay
Problem
Launching GenAI features required early validation with real users. Traditional approach (iterate until 100% perfect) is slow, expensive, risky. Needed structured method to validate early and de-risk investment.
Solution: Beta Program Framework
Structured beta program enabling real customers to test pre-production features through time-limited rollouts. Dual goals: (1) Validate product quality before production, (2) Create innovation communication potential. Designed UX journeys, onboarding flows, and feedback loops with 300+ community members.
Execution & Impact
Strategy & Planning: Coordinated cross-functional teams (Business, Marketing, UX, QA, Legal). Built KPI framework tracking adoption rate, usage frequency, satisfaction, and session metrics. Established technical quality gates (latency, accuracy, 99% safety score).
Pilot Rollout: Launched Voice Stories feature (location-aware AI stories via ChatGPT) across 4 markets (USA, DE+UK, China, APAC). 100 selected customers, <6 months, ~€1M budget. Multi-phase rollout: internal → power users → broader audience with weekly iteration cycles.
Impact: Beta program transformed customers into co-creators. Enabled early go/no-go decisions—collect feedback before full rollout or kill if adoption insufficient, preventing wasted production investment.
Key Learnings
Community over scale: 100 engaged users gave better insights than 1000 passive ones. Quality gates matter: 99% safety score threshold caught issues before user exposure. Reusable framework: Beta structure now template for future GenAI features at Porsche Digital.