Deepful AI
A pharma SaaS platform rebuilt in Framer from a supplied Figma, with a CMS driven blog, Rive animation in place of heavy video, and three languages.
Overview
Deepful is a B2B SaaS platform for pharmaceutical companies, founded by Gustavo Comitre. It serves two distinct groups inside pharma organizations: managers, and the reps and medical science liaisons working in the field.
The company was moving while the site was being rebuilt. It refocused on the European market, reworked its positioning against client feedback, and raised venture funding.
The existing site was slow, weighed down by heavy video, and the decision was to rebuild rather than patch. Shammas Anwer led the Framer build from a Figma design the client supplied.
The challenge
A deadline set by an event, not a preference. The rebuild had to be live before a European roadshow. That fixed the date, and everything else had to fit inside it.
Two audiences on one site. Pharma managers and the reps and MSLs in the field want different things from the same product. The site had to address both without splitting in two.
Legacy URLs that could not move. The privacy policy and terms pages are linked from inside the product, so their paths had to survive the rebuild exactly as they were.
Three languages, and the free tier did not cover it. English, Spanish and Portuguese. Portuguese alone pushed past the five thousand word per locale limit on Framer's free localization.
The old site was slow, and the cause was known. Heavy video files. That ruled out repeating the same approach in the rebuild.
Goals
- Rebuild the site in Framer from the supplied Figma design
- Give the blog a CMS with featured toggles and filtering, so the team can publish without a developer
- Replace the contact page with a demo request that lands where the team already works
- Keep the privacy and terms URLs identical to the old ones
- Ship in English, Spanish and Portuguese
- Be live before the European roadshow
Approach
Rive rather than video. The old site was slow because of heavy video. The rebuild carries its motion as Rive animations instead, including a medicine animation and an animated map, which gives the page movement at a fraction of the weight. Designing out the cause mattered more than matching the old visual effect.
Pay for localization rather than hand build it. A manual language toggle was considered and rejected. Framer's paid localization add on scales as content grows, where a hand built toggle needs maintaining every time a page is added. AI produced the first translation pass and the client's marketing team reviewed the Spanish and Portuguese.
Send demo requests where the team already looks. The contact page was replaced with a popup survey capturing name, email, company and an open details field, wired through to Google Sheets. No new tool for anyone to learn.
Preserve the legacy paths exactly. Privacy and terms were rebuilt at the same URLs the product already links to, so nothing inside the app broke on launch day.
Settle small decisions by building the options. Two preloader versions were built and the one without a preloader was chosen. Four CTA gradient treatments were built and the fourth was chosen. Quicker than describing them.
Fix the terminology. The site said MCLs. The correct term is MSLs, medical science liaisons, and it was corrected everywhere it appeared.
Design
The visual design came from the client as a finished Figma file. The work here was implementation, and the decisions implementation forced.
Inter carries the typography. SF Pro was ruled out because Apple's licensing does not permit it for website use.
Motion runs on Rive rather than video, a medicine animation and an animated map, both chosen to keep the page light on a site whose predecessor was slow.
Build
Built in Framer by Shammas Anwer, from the client's Figma design.
Pages and structures: the landing page, a blog list and a blog detail template running on the CMS with featured and not featured toggles plus filtering, and privacy policy and terms pages rebuilt at their original URLs.
Also implemented: the demo request popup wired to Google Sheets, the Rive animations, three locales through Framer's localization add on, SEO metadata, favicon, social sharing image, and the responsive work across the site.
Support continued after delivery: mobile and iPad responsiveness fixes, a duplicate title on mobile, terms and privacy content added in English and Portuguese, formatting repairs after the client's marketing assistant edited the Portuguese locale, and minor spacing fixes months later.
Outcome
The site went live on 14 August 2025 at deepful.ai, in three languages, ahead of the European roadshow it was built for. The original target was 10 August, so it landed four days past that date and inside the estimated range for the work itself.
One request could not be delivered as asked. The client wanted the focused case study card highlighted in orange with the others greyed back. Because the sticky cards sit inside a single section, that state could not be triggered per card. After several attempts the existing transition was kept, and the people who reviewed it preferred it. That belongs in the record as a platform limit rather than being quietly dropped.
The engagement carried on past the order. Fixes continued for weeks by explicit agreement, the client's marketing analyst was given a draft page and event templates inside the same Framer workspace to build a separate roadshow landing page, and small maintenance requests were still arriving in February 2026.
No traffic, ranking or conversion figures are published here. Deepful has not shared post launch data, and the rebuild was never measured against the site it replaced.
Key takeaways
- A date set by an event is a different kind of constraint from a preferred date. Scoping backwards from the roadshow decided what got built and what waited.
- When a rebuild exists because the old site was slow, the cause has to be designed out, not reproduced. Video came out and Rive went in.
- Some requests are platform limits rather than effort problems. The per card highlight could not be triggered from inside one sticky section, and saying so was worth more than further attempts.
See it in the browser
The finished site is live. Open it and judge the work in its real context.
Have a project in mind?
Whether you're launching something new or improving an existing website, let's discuss your goals and see if we're a good fit.
