Every company below comes from a published Filestack case study, and each section links to its source. Customer claims, quotes, and results stay within what those sources document. Where a request or image was made internally to show the mechanism, we label it as a Filestack example rather than customer output.
The pattern running through all ten is the same, and it is not really about files. Each of these teams decided that uploads were not the thing they were in business to build.
Key takeaways
- Grailed and Pingboard replaced upload systems they no longer wanted to maintain.
- BuildZoom and HeyOrca use one stored image to serve different sizes at delivery.
- One Hour Translation removed an upload ceiling and reported that upload support calls fell to zero.
- Sweeps and Imagine Learning Classroom use browser capture to make uploaded evidence part of the product.
- Keela compared Filestack with the time and cost of building file processing in house.
1. Grailed replaced uploads it was maintaining itself
Grailed is a peer to peer marketplace for high-end menswear. CTO Julian Connor calls it an image heavy app where “we handle millions of images a day.”
What they run: the File Picker, taking files from a local drive, social media or a user’s webcam. Before that, Grailed managed uploads itself, which Connor calls “the largest bottleneck in our entire app.” The case study notes they did a custom integration rather than a drop-in one.
2. BuildZoom serves one original at every size
BuildZoom connects property owners with contractors, and galleries of finished projects are the product. Artem Kholodenko, their Director of Front-End Engineering, gave the plainest reason in the whole set: “We’re not in the business of building file uploaders.”
What they run: the Filestack Rails gem for upload, Amazon S3 for storage, and the Image Transformations API at delivery. Their case study is specific about why that last part matters. BuildZoom’s database of original high resolution images is “served only as big as they need to be to ensure a minimal page footprint”, and they credit that with better page rankings as a side effect.
The case study describes that delivery pattern but does not publish these exact URLs. This internal Filestack example shows a gallery thumbnail and detail view from the same stored handle:
https://cdn.filestackcontent.com/resize=width:320/compress/output=format:webp/HANDLE
https://cdn.filestackcontent.com/resize=width:1400/compress/output=format:webp/HANDLEResize first, then compress, then encode, so the format conversion runs on a small image rather than a large one.
3. One Hour Translation removed a 30 megabyte ceiling
One Hour Translation runs professional human translation in over 75 languages with 15,000 translators across more than 100 countries. CTO and co-founder Oren Yagev is direct about the reason they looked: “We had a limit of 30 megabytes and had issues because of that.”
What they run: uploads on the JavaScript side, with files delivered straight into their own S3 bucket, which Yagev describes as “like a glove and hand.” They chose Filestack over one other candidate “because it’s the most documented.”
The result is the headline on their case study, which is that upload support calls went to zero. In Yagev’s words, “since we integrated with Filestack, we don’t hear anything from our customers about uploads.”
4. Sweeps ships one upload flow across three platforms
Sweeps is a labor marketplace for moving and odd jobs, serving over 5,000 workers and more than 10,000 customers. Founder Morris Gelblum describes the problem before Filestack as trust rather than technology: “Lack of trust without profile photos, timeclock photos, etc.”
What they run: the File Picker inside a Ruby on Rails application, with Turbo on iOS and Android, so one integration covers all three platforms. Gelblum lists the deciding factors as “Price, developer ease, multiplatform.”
Their reported results are a near 100 percent success rate on profile photos and timeclock photos becoming a requirement, which is what closed the on-site verification gap. Gelblum also describes the scope growing on its own: “We realize more and more of our system can be done via users taking photos and uploading documents.”
5. Pingboard normalises photos it did not receive
Pingboard builds org charts, and founder and CTO Rob Eanes had written enough upload code to stop wanting to. “We could have built our own image workflows, but it probably wouldn’t have been efficient and probably wouldn’t have been on the right architecture for such a specialized task.”
What they run: the most specific pipeline in this list. Employees upload a local file or connect Facebook, Twitter or Instagram. Then, in the case study’s words, “By starting with Filestack’s Facial Detection capability, Pingboard has built a workflow that crops, resizes, and adds a dynamic background.”
The case study documents that sequence but does not publish this exact request or image. This internal Filestack example uses a sample portrait to show the same processing order:
https://cdn.filestackcontent.com/crop_faces=faces:1,mode:thumb/resize=width:240/6KHnf6yQmuNZPbxpaH2AThe sample output is:
The detail worth stealing is that every picture is normalised through the processing engine “even if they weren’t uploaded using Filestack.” The transformation layer is doing work on files that arrived by other routes entirely.
6. Portfolium stopped pushing straight to S3
Portfolium gives students digital portfolios so hiring managers see work rather than bullet points. CTO and co-founder Daniel Marashlian had met the problem before: “I know the pain your engineering team goes through every day, of not only photo delivery, but photo processing.”
What they run: upload, processing and distribution as one layer, replacing direct-to-S3 uploads. “We were pushing files straight to S3, and it was like, there has to be a better way to upload and maintain them, but also process them. We want to transform them, we want to stream them, we want to eventually distribute them around the world.”
They integrated at 50,000 users. The case study notes they now have four million.
7. Keela costed the alternative
Keela is a nonprofit management platform, and theirs is the case study with a number attached to the decision rather than the outcome. Filestack saved Keela an estimated $25,000 CAD during implementation alone, not counting upkeep, against in-house developers and at least three months for a first version.
What they run: file processing as a bought component alongside Mailchimp, Sendgrid, Stripe and PayPal. Co-founder and CEO Nejeed Kassam describes the stack as bringing in “the best players to run the play.” He is blunt about the alternative: “it would have been a stupid amount of work to build it from scratch.”
8. HeyOrca pulls from the tools its clients already use
HeyOrca helps marketing agencies plan and approve social content. They tried DIY solutions and jQuery plugins first, and none of it held.
What they run: three parts of the platform. The File Picker gives clients access to their cloud drives and took “only a few lines of JavaScript” to integrate. Dynamic resizing serves one stored image at different sizes by changing resize in the URL. Video transcoding through the REST API converts content into web-friendly formats.
Their quote on commercial impact is unusually direct: “Without the Google Drive and Dropbox integrations through Filestack, we wouldn’t get the agency businesses we do.”
9. Imagine Learning Classroom records in the browser
Imagine Learning Classroom, formerly LearnZillion, is a K-12 lesson player, and the integration sits in lesson authoring.
What they run: the File Picker for office files, direct uploads from Google Workspace and Drive, and a browser-based slide editor for images and audio. An audio overlay lets educators narrate a slide by recording in the browser. The case study says that capability mattered during the pandemic, when remote teaching required narration on existing lessons.
10. Classcard found the document viewer
Classcard is a platform for class providers, mostly children’s classes and fitness programs. They evaluated several options including Uppy before choosing, on documentation quality, the managed approach and ease of implementation.
What they run: uploads, CDN delivery, and the document viewer. The viewer lets providers show materials inside Classcard and became an unexpected “aha” moment for the company and its users. Many users are not technically confident, which made the picker interface decisive. Their representative sums up the result: “The entire file management process is now taken care of with Filestack.”
What the ten have in common
Three things, and none of them is a feature.
Uploads became a maintained system before anyone decided to buy. Grailed called it their largest bottleneck. Pingboard had already built several and liked none of them. Portfolium was pushing straight to S3 and knew there had to be a better way. Keela priced the alternative at $25,000 and three months.
The visible product is almost never the upload. It is a verified worker, a face on an org chart, a narrated lesson, or a fast portfolio. Users notice what happens after the file arrives. Six of these ten companies describe transformation or delivery as the part that mattered. The operations these builds use are all available to try on a free key today.
The transformation layer also handles files that arrived by other routes. Pingboard normalises pictures that were not uploaded through Filestack. BuildZoom keeps one high-resolution original and sizes it per context. HeyOrca serves one stored image at any size by changing the URL. The internal sample portrait used above also serves as a compact WebP thumbnail from the same handle.
The image editing api guide covers those operations. Our format guide explains when to convert to webp, and the file delivery walkthrough covers caching.
What the free plan covers of this, and where it stops
Most of the mechanisms above run on a free key. Of 65 operations tested in July 2026, 50 work on the free plan, including resize, compress, format conversion, caching and the facial detection family that sits behind the Pingboard pattern.
Thirteen return a 403 and belong to the higher plans: smart_crop, enhance, upscale, redeye, tags, sfw, caption, ocr, copyright, doc_detection, image_sentiment, text_sentiment, and video_convert. The line is consistent. If an operation has to understand what is inside a file, it is on a higher plan. If it changes the file’s shape, size or format, it is on the free one.
That matters for two of the companies above. HeyOrca’s video transcoding is a paid operation, and any workflow that needs tagging or moderation rather than face detection lands on the same side. The pricing page shows which plan carries each one.
Which of these is your situation
A more useful question than which company you resemble is which sentence you have already said out loud. Grailed said uploads were their biggest bottleneck. Pingboard said the architecture was wrong for a specialised task. One Hour Translation said they did not want to be an upload company. Portfolium said there had to be a better way. Keela said building it would have been a stupid amount of work.
If one of those is familiar, the shortest route to knowing is to build the smallest version of the relevant pattern and see how far it gets. Pingboard’s documented pattern is the most self-contained one to reproduce internally: detect a face, crop around it, and normalise the result. That tests the same processing sequence without claiming to reproduce Pingboard’s implementation. The face detection side of it is covered in blur faces, and creating an account takes about a minute.
For how these separate calls become a pipeline rather than a set of examples, the file delivery workflow guide covers the chain end to end.
Joshua is a web developer with over 4 years of experience building responsive, high-performance websites and web applications. Currently working as an AI Automation Specialist, he combines modern web development with automation to create efficient, scalable digital solutions. He shares practical insights on WordPress, web development, and emerging technologies.
Read More →