Sumo AI

Overview

Sumo AI offers 500,000 hours of real human conversation, sourced from the U.S. legal services industry and pitched as the largest commercially available dataset of its kind. This AI training data landing page turns a dense, technical product into a clear catalogue page that machine-learning teams can evaluate quickly. I designed and built it in Framer for the client.

The challenge

Buyers of training data are technical and sceptical. Before they get in touch, they want to know what is in the dataset, how it was transcribed and annotated, what they could build with it, and whether it is compliant. That is a lot of information, and in most data listings it ends up as an unreadable spec sheet. Sumo AI needed a page that answered those questions in a logical order without overwhelming the reader.

What I built

• A hero that leads with the dataset's scale and its legal-services source
• Four headline qualities: human-verified transcripts, real-world (not scripted or synthetic) calls, included metadata and legal-domain depth
• A use-case section covering noisy-condition speech recognition, emotion and tone recognition, indirect speech, turn-taking, code-switching and noisy transcript comprehension
• Tags on each use case so researchers can spot their field at a glance
• A technical specs block covering audio and text formats, native annotations and optional human annotations
• A compliance section explaining de-identification and PII redaction

Design & build

The direction is quiet and editorial: an earthy palette, crisp type and a strong grid so detailed information stays easy to scan. Each section answers one question a buyer is likely to ask, in the order they would ask it, moving from "what is it?" to "what can I build?" to "is it safe to use?". The build is in Framer, which keeps the page fast and lets the team update specs and use cases as the catalogue grows. It works like a small web app for evaluation rather than a brochure.

What it delivers

Sumo AI now has a product page that does the first round of qualification on its own. Technical buyers can understand the dataset's contents, quality and compliance without a sales call, and arrive at the conversation already convinced it fits their use case. The clear structure makes a complex offer feel trustworthy, and it gives the team a reusable template for presenting future datasets.