Zapigo

Zapigo started with a genuinely interesting inversion on gifting: instead of the giver choosing a gift and hoping it lands, the receiver gets to pick from a curated assortment. Less guessing, more delight.

We launched with books. I designed and built the entire UI from scratch — the catalog browsing experience, the gift selection flow, the card reveal moment. End-to-end, with no existing design system to lean on.

A gifting platform lives and dies by its catalog. I built the AI ingestion and tagging pipelines that processed over one lakh products — classifying items as giftable, enriching metadata, and tagging by category, occasion, and recipient type.

Beyond software, I also led the physical gift card design: how the unboxing experience looked and felt, and how the card communicated the receiver-choice mechanic clearly without instructions.

• 100,000+ products ingested and enriched through Langchain-based pipelines
• Automated giftability classification and multi-tag enrichment
• Physical card design tying the digital selection flow to the unboxing moment
• Brand partnerships onboarded — coordinating catalog integration for each brand

The gifting model did not get the traction we needed. We stepped back, looked at where the real friction was, and pivoted toward digital invites and guest management — a more frequent use case closely adjacent to the occasions gifting was built around.

I built the MVP. The backend runs on FastAPI with a PostgreSQL database. The first frontend was Flutter and FlutterFlow for rapid prototyping; as the product matured, we migrated to Next.js for the production frontend.

One of the more interesting technical problems was building a hybrid search that combined pgvector-powered semantic search with traditional Postgres filtering. Pure keyword search was too brittle; pure vector search missed structured filters. The hybrid approach made both work together cleanly.

The whole stack ran on Azure — blob storage, compute, managed Postgres. As a founding team member, I wore the cloud infra hat too: provisioning, environment management, and keeping costs sane as the catalog grew.

This was the full founding team experience: product decisions with no playbook, pivots that required throwing away months of work, and the satisfaction of shipping things that actually ran in production.

Across design, frontend, backend, AI pipelines, brand ops, and infrastructure — the role was less a job title and more a commitment to making the product real, whatever that required on any given week.