Hi, I'm
Product consultant by day, AI builder on the side. I've shipped 5 AI products, led enterprise rollouts across 72 business units, and I'm figuring out the AI + Product space — honestly, and in public.
I started as a consultant at Bosch who kept asking "why are we building this?" instead of just implementing requirements. That question followed me through an MBA, into IBM, and now into everything I build.
At IBM, I worked on Oracle Expenses for 72 business units globally — including navigating a China Fapiao integration that needed a custom approach beyond Oracle's standard setup. On the GenAI side, our team found that fixing hallucinations in a test script agent wasn't about model parameters — it was about feeding it better business context. Small, collaborative efforts across the team moved accuracy from unusable to 60-70%.
Along with work, I build AI products. A content curation PWA backed by neuroscience research. A meal companion for solo diners. An ATS resume optimizer I use for my own applications. A RAG chatbot for Oracle docs. A productivity app with swipe-based logging. These didn't start in a vacuum — they came from real problems I saw around me, from family, friends, and colleagues — and nothing out there solved them well enough.
I believe as AI makes execution cheaper, the most valuable skill becomes knowing what's worth building. That's the space I want to work in.
Anti-Content-Bubble PWA
Subscription app that fights algorithmic content bubbles with daily AI-curated reads, progressively expanding your intellectual world.
I kept ending my days exhausted but couldn't figure out where my time actually went. Spreadsheets, to-do apps, Airtable — all had the same problem. Logging felt like a chore, so I'd quit after 2-3 days. DayFlow replaces rigid forms with swipe cards and physics-based bubble inputs, tracks mood and energy alongside activities, and sends an AI coaching summary each night based on patterns it finds.
At IBM, colleagues spent hours searching through 1000+ pages of Oracle Fusion PDFs for navigation paths and configuration steps. Google and ChatGPT gave generic answers that missed Oracle-specific terminology. This RAG chatbot ingests 11 curated documents (~37.7 MB), provides verified answers with exact source citations, and refuses to guess when the information isn't in the docs.
75% of resumes get filtered out by ATS systems before a human ever sees them. I was applying for AI PM roles and needed a tool that could extract job-specific keywords, restructure bullets using the STAR method, and strip out AI-sounding language — without paying $20-50 per optimization. Built a multi-LLM tool supporting Claude, GPT-4, Gemini, and OpenRouter with a bring-your-own-key model. Zero hosting costs, zero data stored.
My sister cooks alone in Bangalore — every recipe app defaults to 4 servings, wastes half her groceries, and ignores what's already in her fridge. SoloBite is an AI meal companion built for solo diners: snap your fridge, type what you have, or say what you're craving. You get a single-portion, nutritionally-complete recipe with exact macros and a video walkthrough. No accounts, no data stored.
My husband noticed his YouTube feed had become an echo chamber — same topics, same perspectives, every day. Research shows unfamiliar content activates different brain regions, but no product builds around that insight. Debubble is a subscription PWA that fights content bubbles with daily AI-curated reads, progressively expanding your intellectual world through increasingly unfamiliar but connected topics.