fruitpunch

Guide · AI & data

FruitPunch AI: History of the Eindhoven AI for Good Community

By the FruitPunch research deskUpdated November 15, 20269 min read

In short

FruitPunch AI was a community and later a company founded by TU Eindhoven student Buster Franken in September 2018, known for AI Connect events and 10-week AI for Good challenges with partners such as Cornell, Colgate University and Rijkswaterstaat. Its 2023 website claimed 1,500+ AI engineers, 19 challenges and 50+ partners, and a 2024 IO+ interview reported 4,500 trainees and 38 challenges. The current fruitpunch.ai is an independent site with no link to that organization.

FruitPunch AI was an Eindhoven, Netherlands, community that taught people to apply artificial intelligence by working on real problems for conservation groups, agencies and researchers. It started at Eindhoven University of Technology (TU/e): the university magazine Cursor reported that it was founded "as recently as September 2018" by student Buster Franken. It grew from campus events into a company that ran 10-week "AI for Good" challenges, and by April 2024 an IO+ interview reported that Franken and co-founder Sako Arts had 4,500 trainees and 38 challenges behind them.

A note on this website: the fruitpunch.ai domain now hosts FruitPunch, an independent rankings and guides site. It is not FruitPunch AI, is not affiliated with it or its founders, and does not run challenges, courses or certificates. This page is a neutral, sourced account of a past project, written for readers who arrive from old links. Quotes and figures attributed to FruitPunch AI come from its own archived pages in the Internet Archive's Wayback Machine or from press coverage.

Key facts

  • Founded: September 2018 at TU/e by Buster Franken, with Mickey Beurskens as an early co-founder (Cursor, 2019).
  • First big event: AI Connect on December 12, 2018, at TU/e Innovation Space, with IBM among the partners (archived event page), attended by about 150 people (TU/e).
  • Largest AI Connect: June 15, 2019, with more than 300 attendees (TU/e).
  • Challenge format: about 12 hours a week for 10 weeks, according to its archived challenge pages (AI for Seals).
  • Funding: LUMO Labs Fund II invested in January 2021 (High Tech Campus Eindhoven), followed by a €500,000 round led by Shamrock Ventures in May 2023 (IO+).
  • Reported size: 1,500+ AI engineers, 19 challenges, 3 chapters and 50+ partners on its 2023 about page (archived).

How did FruitPunch AI start?

The founding story appears in several versions, and they agree on the core. Cursor wrote in February 2019 that Franken, then a pre-master student in data science, looked for an AI community in Eindhoven, found none and founded one. Franken told the magazine there was no such community "in the Brainport region", the name used for the Eindhoven technology cluster. Cursor also reported that the team then had six members from different faculties and worked with Data Science Center Eindhoven (DSC/e) and IBM.

FruitPunch AI's own 2021 about page told the name story: the founders took part in RoboCup in Montreal, were offered fruit punch by people who ran e-sports competitions, and decided their first event would be an AI e-sports competition. Innovation Origins covered that plan in January 2019, reporting that teams from Eindhoven, Tilburg and Utrecht had signed up and that IBM would provide servers. Franken described the competition as "a kind of RoboCup for AI, with Eindhoven as the epicentre."

By June 2019 Innovation Origins described FruitPunch AI as a student team, an early start-up and a foundation, with "6 full-timers, 6 part-timers and many dozens of some-timers". The 2021 about page gave its contact address as the Gas lab at TU/e (archived), and the 2024 IO+ interview took place at the Microlab on Strijp-S in Eindhoven (IO+).

What were AI Connect and AI Code?

These were the two event formats from the campus years.

AI Connect was a networking and talks evening. The archived page for the first edition lists December 12, 2018, at TU/e Innovation Space, speakers from IBM and TU/e, and partners TU/e Innovation Space, DSC/e, the Eindhoven Data Science meetup and IBM (archived). TU/e said about 150 researchers, students and professionals attended (TU/e). The June 15, 2019 edition drew more than 300 people, with talks on applied AI ethics, evolutionary computing and swarm intelligence and an AI market where companies showed projects (TU/e).

AI Code was a recurring coding session. In June 2019 Franken mentioned a code night "every Tuesday evening" (Innovation Origins). The archived session pages are numbered as "epochs"; epoch 0 covered implementing linear regression with gradient descent (archived), and a later page announced that sessions would restart on February 23 (archived).

How did the AI for Good challenges work?

By 2020 the focus had moved to challenges. The 2023 about page says the first one came after South African wildlife reserves answered a call for AI projects tied to the UN Sustainable Development Goals: Franken visited Pilanesberg, then recruited Sako Arts and 50 AI engineers who spent 10 weeks building an autonomous drone with thermal cameras to detect poachers. The 2024 IO+ interview gives a different head count for the same project: a LinkedIn call "resulted in a group of eighty AI engineers" (IO+). Brainport Eindhoven also profiled this anti-poaching drone work (Brainport Eindhoven).

The model that followed had a few fixed parts, based on the archived pages:

  • A partner with a real problem and data. Partners named on challenge pages include Colgate University, Rijkswaterstaat (the Dutch public works agency) and ReefSupport.
  • A time box. Pages asked for about 12 hours a week over 10 weeks (AI for Coral Reefs). The 2024 interview put the range at eight to ten weeks.
  • Mixed teams. The coral reef page said selection looked for "a combination of rookies looking to learn, hardened professionals and life-long learners" (archived).
  • Labs. By late 2023 the site grouped work into an AI for Wildlife Lab, an AI for Earth Lab and an AI for Health Lab (archived).

The business side, as High Tech Campus Eindhoven described it in 2021, was that revenue would come from organizations that needed AI talent or solutions, while community members got free education (High Tech Campus Eindhoven).

Which challenges did FruitPunch AI run?

The table lists six challenges named in press coverage and on archived pages. Dates come from the archived pages and show when each was announced or scheduled.

Challenge Period on archived page Partner named Task described
AI for Coral Reefs Late 2021 ReefSupport Detect and classify coral in dive survey images
AI against Oil Spills August to October 2022 Rijkswaterstaat Detect oil spills in satellite and drone images
AI for Seals December 2022 to early 2023 Colgate University Improve SealNet facial recognition for harbor seals
AI for Forest Elephants Early 2023 Cornell, Stanford Detect elephant rumbles and gunshots in audio
AI for Bears Early 2024 Not named in text Detect and identify bears on low-power hardware in Romania
AI for Nepali Farmers Early 2024 Not named in text Read Nepali loan records with OCR

Some details from each page:

  • Coral reefs. The page claimed that "for every 4 hours of diving, the analysis of recorded material takes 5 days" and quoted the founder of the Indonesia Biru Foundation (archived).
  • Oil spills. The goal was "automating detection and classification of oil spills in satellite data and in drone images on edge", with a focus on small spills in ports and inland waterways that satellite monitoring missed (archived). A second oil spill challenge page was archived in 2024 (archived).
  • Seals. Colgate University's SealNet was described as "the first marine mammal facial recognition software", and the stated goal was to increase its accuracy and robustness (archived).
  • Forest elephants. The page named the Cornell Conservation Bioacoustics Center and Stanford University and described recordings from acoustic units in the Central African rainforest, with a target of microphones running "for up to 4 months" on batteries (archived).
  • Bears. The stated aim was to cut human and bear conflicts "in the five communes of Romania by 50%" with a real-time detector that triggers repellents (archived).
  • Nepali farmers. The task was to extract text from loan contracts in Devanagari script and translate it into English for agricultural cooperatives (archived). IO+ described this as helping women farmers in Nepal by digitizing loans (IO+).

The 2023 about page also mentioned work with European medical centers on COVID detection and with the European Space Agency on wildfires and illegal deforestation (archived).

How big did the community get?

The reported numbers rose over time, and they come from the organization or from interviews rather than an independent count.

Date Figure Source
February 2019 6 team members Cursor
June 2019 6 full-timers, 6 part-timers and "many dozens" of occasional helpers Innovation Origins
December 2023 1,500+ AI engineers, 19 challenges, 3 chapters, 50+ partners Archived about page
April 2024 4,500 trainees, 38 challenges, over €1 million invested IO+

On money, the 2021 about page said Franken attracted "half a million in seed investment from LUMO Labs", while High Tech Campus Eindhoven reported the LUMO Labs Fund II investment without an amount (High Tech Campus Eindhoven). The May 2023 round was €500,000, led by Shamrock Ventures with LUMO Labs and angel investors, and IO+ named NXP, Stanford University and WWF as organizations the company supplied AI engineering work to (IO+).

What did the press write about FruitPunch AI?

Which open projects and datasets relate to these challenges?

Several challenges built on published research tools that anyone can read or reuse:

  • SealNet (Colgate University): Birenbaum and colleagues describe face detection and recognition for harbor seals in Ecology and Evolution, 2022, reporting 88% Rank-1 accuracy on 1,752 images of 408 seals.
  • BearID: Clapham and colleagues trained a brown bear face recognizer on 4,674 images of 132 known bears, reaching 83.9% identification accuracy (Ecology and Evolution, 2020). The AI for Bears page listed BearID data among its datasets. A follow-up paper tested the method across bear species (Mammalian Biology, 2022).
  • Forest elephant acoustics: Bjorck and colleagues, with Cornell's Peter Wrege, describe detection and compression for passive acoustic monitoring of the African forest elephant (arXiv, 2019).
  • MegaDetector: Microsoft's open camera trap model, now listed in its biodiversity hub on GitHub. The AI for Bears page mentions a dataset with "MD-generated bounding boxes" (archived).

Frequently asked questions

Is the current fruitpunch.ai the same as FruitPunch AI?

No. The domain now hosts FruitPunch, an independent rankings and guides site. It is not affiliated with FruitPunch AI, its founders, partners or members, and it does not run challenges or issue certificates.

When was FruitPunch AI founded?

Cursor reported a founding date of September 2018 (Cursor). The organization's own pages place the idea in summer 2018 (archived), and IO+ says it became a company by the end of 2020 after operating as a foundation (IO+).

Who founded FruitPunch AI?

Buster Franken founded it as a TU/e student, first with Mickey Beurskens (Cursor). Sako Arts later became co-founder and CTO (archived).

How long was a FruitPunch AI challenge?

Archived challenge pages describe about 12 hours a week for 10 weeks (archived); a 2024 interview said eight to ten weeks (Brainport Eindhoven).

How many people took part?

The organization claimed 1,500+ AI engineers in 2023 (archived), and IO+ reported 4,500 trainees and 38 challenges in 2024 (IO+). Neither figure was independently audited.

Where can I read the original FruitPunch AI pages?

The Internet Archive's Wayback Machine holds copies, including the 2021 about page and the 2023 about page.

Spotted an error in this guide? Email editorial@fruitpunch.ai

Keep reading