Data Platform — from 0 to 1 in 6 months.
How a two-person product design team shipped an AI-powered SaaS platform end-to-end — by collapsing the line between Product, Design, and Engineering.
A new operating system for fund data.
Data Platform turns the messy, document-heavy world of private fund operations into a single, auditable workflow. AI-assisted ingestion, extraction, validation, and delivery — built from scratch, with engineering and product moving in lockstep.
Design at scale.
Volume, breadth, and depth — designed in parallel, not in sequence.
In 6 months, one design resource took Data Platform from concept to launch-ready — covering 12+ design areas and 50+ UX flows. Here's a snapshot of the scope.
A new way to design product.
Linear no longer. Product, Design, and Engineering — in sync, not in sequence.
We started with Lovable — which let our cross-functional team align on vision, pressure test backend architecture, and jump straight into ideation, all at the same time. No waiting. No handoffs. Then came Figma Make, Claude, and a tight loop of real-time client feedback. We used AI tools to brainstorm efficiently and at speed, sometimes diving into original Lovable prototype to granualarly define UX flow and interactions and sometimes ideating requirements using Claude (artifact) What would've taken months compressed into days. We went from 0 to 1 while simultaneously getting the product in front of real users and iterating live.
The discover-ideate-design-test-iterate cycle? Still there. But it's not linear anymore. Product, Engineering, and Design working together in agentic ways — that's the shift. This is what the future of product design looks like. Product, Design, Engineering — not in sequence, but in sync.
This is more than a launch. It's a glimpse into the future of AI-empowered, collaborative product development.
The shift
Sequential. Slow. Siloed.
Simultaneous. Looped. In sync.
Lovable, on day one. Before a single Figma frame existed, the team aligned on vision and pressure-tested backend architecture inside a working Lovable prototype — Product, Design, and Engineering ideating on the same artifact, in the same room, at the same time.
Designed with our clients in the room.
Instead of waiting for a full production release, we fast-tracked the work — shipping modules to prod in phases, putting them in front of real clients, and rebuilding the loop with their feedback in days, not quarters.
We didn't wait for "done" to learn. Two modules were shipped to production in phases during the beta release, and Product and Design jumped on calls with eight design-partner clients — leading them through real tasks, watching where they stumbled, capturing what surprised them. Every session ended with a written summary back to the client: what we heard, what we're fixing now, what's queued, what we're still evaluating.
Then we moved. Partnered tightly with our UI engineer, the team turned same-week sessions into same-week fixes — column overlap bugs, fund-vehicle dropdowns, AI-search guidance, modal sizing, drag-and-drop scope. Heavier items — Excel native support, persistent field selections, granular source attribution — flowed into the backlog with their provenance attached.
This is what the loop looks like when it's real. Not research-as-ceremony — research as a continuous obligation to the people using the product.
Testing Waves by Module
Four of eight design partners shown. Sessions and follow-ups tracked in our User Testing Feedback DMP mailbox — each one closed with a structured summary covering what worked, what didn't, what we fixed, and what's next.
"What is working well: the AskDocuments Conversational AI was well-received, with the team seeing clear value for business intelligence queries and summaries across the document corpus. Manual document upload is straightforward, and the upload modal interaction was completed quickly and successfully." — UC Regents Design Partnership Session, May 2026
The toolchain that made it possible.
Each tool plays a distinct role — together they collapse the gap between idea and product.
What changed.
Not just what we shipped — how we shipped it.
Designed for AI, end to end.
Designing AI experiences is its own discipline. Across the platform we shipped three flagship AI flows that show how machine intelligence can disappear into the workflow rather than sit on top of it.
Conversational AI Chatbot.
An always-available AskDocuments assistant that lives in the app shell. Users ask in natural language — summarize Q4 docs, extract capital call dates, surface duplicates — and get cited, structured answers grounded in their own documents.
AI Search & Filtering.
A unified search bar that interprets intent in plain English and translates it into faceted filters across hundreds of fund documents — designed to feel like asking a colleague, not querying a database.
AI Insights on Document Manager.
Inline, on-demand intelligence over any document set — comparison reports, trend spotting, exception summaries — generated from the documents already stored in the system. AI shows up where the work is, not in a separate tab.
Ingestion, shipped. The final design of the document workflow's first step — clean upload & integration setup at the top, an AI-powered search and filter bar over the ingested-documents table, and live status pills (Validation Required, Queued, Ingested, Duplicate, Storage Only) that come from the extraction and validation engines.
From sidepanel to full report. Conversational AI scales gracefully — a quick question gets a quick answer in the side panel; a complex prompt expands into a full, cited report that lives inside the platform. The same surface, three different intensities of work.