👋🏼 Hey There, I’m Rohan Shah!
As a founding Product Manager at Outbox Labs, I had the freedom to experiment, learn from quick failures, and build successful products. Our team always focused on shipping fast, and using AI was the best way to reach that goal. I use AI every day for planning and prototyping, but the projects below are my most important wins. They highlight my journey in building AI-first products that solve real problems.
ReachInbox.ai is an AI-driven cold email outreach platform designed to automate the entire sales prospecting cycle. It focuses on helping businesses find high-intent leads, craft personalized message sequences, and maintain high deliverability so that emails consistently land in the recipient's primary inbox rather than the spam folder.
This was our flagship product and I was responsible for the entire lifecycle of this product. Below are some of the biggest AI Features and projects where AI Product Management helped fast track development significantly
<aside> 💡
TL;DR
The Problem: Unstructured lead data made traditional filtering inaccurate and frustrating for users.
The AI Solution: Integrated GPT-4o mini to transform natural language prompts into precise database queries.
Key Achievement: Replaced static roles with semantic matching, increasing search accuracy by 60% and converting 5k users to the new subscription tier. </aside>
Click here for detailed explanation
<aside> 💡
TL;DR
The Problem: Incomplete or dirty lead profiles were causing a drop in engagement and broken email templates.
The AI Solution: Built an orchestration layer using Firecrawl, IcyPeas/Operia and GPT-4o mini for automated website/LinkedIn scraping and data cleaning.
Key Achievement: Optimized token costs via schema-filtering, resulting in a 250% increase in AI credit utilization and a boost in positive reply rates. </aside>
Click here for detailed explanation
<aside> 💡
TL;DR
The Problem: Standard spam checkers broke the human tone of emails and failed to provide contextually accurate replacements.
The AI Solution: Developed a hybrid system using NLP techniques like stemming and lemmatization combined with GPT-4o mini for contextual re-writing.
Key Achievement: Pivoted the architecture after AI Evals showed RAG was underperforming, successfully reducing support tickets and protecting IP reputation. </aside>
Click here for detailed explanation
<aside> 💡
TL;DR
The Problem: Sales teams struggled to manage high volumes of incoming replies, leading to delayed follow-ups and lost deals.
The AI Solution: Planned and executed a RAG pipeline using vector embeddings and cosine similarity to retrieve business-specific context for automated drafting.
Key Achievement: Benchmarked multiple LLMs to select Gemini 2.5 Flash for optimal latency, establishing ReachInbox as an industry first-mover in the AI agent space. </aside>
Click here for detailed explanation