Building the brand, website, and launch system for a $12.5M seed AI company.

Datawizz is a San Francisco AI company founded by Iddo Gino, who built RapidAPI before it. Its platform replaces large, general-purpose language models with smaller Specialized Language Models trained for each customer’s own tasks.

It profiles a product’s AI requests, trains expert models on them and routes every request to the model that fits. In September 2025 Datawizz announced a $12.5M seed round led by Human Capital. We worked with the team for more than 18 months, from the brand and the website to investor materials, event and marketing assets and launch visuals.

Client

Datawizz

Industry

AI infrastructure, B2B

Services

Brand identity · Website design · Framer development · Investor materials · Event and marketing collateral · Launch videos

Timeline

2024 to 2026

Datawizz’s product is new to most buyers: smaller models trained on a company’s own requests, in place of the large ones it already uses.

01

Credible from day one

The identity had to read as serious infrastructure to engineers and to investors.

02

A new idea to explain

Specialized models, distillation and routing needed a website that explains them in plain steps.

03

Room to grow

The brand had to stretch from an app icon and stickers to a conference booth, investor materials and launch videos.

We started with the brand and explored six directions before choosing one.

Datawizz chose a bold D with a four-point spark cut out of it, set beside a rounded wordmark on deep black with violet light.

The spark carries through everything that followed: icons, stickers, the booth and the website.

Then we put the brand to work for events and launches: a conference booth, holographic stickers that say Own your AI, and glowing product icons.

The investor materials and launch videos use the same mark, the same deep black and the same violet light.

The homepage explains the platform step by step, starting with Integrate: Datawizz goes into an existing app with no code changes. A comparison then sets specialized models against the large ones teams already use.

We designed it dark, with soft violet and red light behind each section, and built it in Framer. It scores 99/100 on Google PageSpeed Insights.

Each feature got its own section and its own visual, so a technical page never turns into a wall of text.

Pixel art for the energy smaller models save, a log table for inference data, a world map for the global network, a price table per base model and benchmark chips for evaluation.

Product pages go deeper into single features: knowledge distillation, smart routing between models, guardrails against abuse and prompt injection, and pricing for the platform.

Each one follows the homepage: a dark section, one short heading and one diagram that shows the feature.