February 3, 2023
7 min read

Giskard's retrospective of 2022... And a look into what's coming in 2023!

Giskard's retrospective of 2022, covering people, company, customers and product news, and a look into what's next for 2023. 2022 was a pivotal year, as we went from 3 to 10 people, raised our first round, expanded our product and grew our customer base. We share a special announcement, and unveil the key features that will come in Giskard 2.0 this year.

Giskard 2022 in Review: Agenda
Alex Combessie
Giskard 2022 in Review: Agenda
Giskard 2022 in Review: Agenda

🍿 Full video

You will find below the full transcript of the video:

Hello everyone, this is Alex Combessie from Giskard, and today I'll present to you a retrospective of 2022.

I'll be switching to a presentation to have some written help. I'm really pleased to be with you today. For those who don't know us, Giskard, we are building the first collaborative and open-source software to ensure the quality of AI models.

A little bit of background about us: I'm Alex, I'm co-founder and CEO of the company. Before that, I was at Dataiku, so was Andrei, our CTO, who also worked previously at CERN in Switzerland, and Jean-Marie John-Mathews, our Chief Product Officer, who studied AI Ethics during his Ph.D. and was a former Data Scientist from Thales and Capgemini.

Giskard's founders

I'll talk about four points to summarize 2022: the people, the company, our customers and our product. And finally I'll give you an overview of what's next in 2023.

🫶 People

In January 2022, this was a a picture of us, the three founders. That was it, that was the whole company. We didn't have an office, this photo was from the company headquarters, in my apartment. And we had an ugly robot mascot. It was literally generated by AI, this little green robot. We paid 50 bucks to get it from a pre-rendered graphics website.

Giskard founders' fondue night in 2022, with our previous green robot mascot

2023 and here we are, we are a team of 10:

Giskard's whole team in Q1 2023

We are all at Station F, which is the world's largest startup campus, right in Paris. And we have a turtle mascot, so the turtle represents longevity and resilience, values which are super important in these uncertain times.

I want to highlight that we are still hiring: gisk.ar/jobs. If you're interested to join our mission to Quality for AI, we're hiring across the board, from software engineers to Machine Learning researchers and developer advocates.

🏦 Company

Now speaking about the company, and more specifically its finances.

January 2022 was really tough. I took a screenshot of a Slack notification from our bank. At a point in time, we had only 700 euros left in the bank. We had to stop all expenses. We genuinely, I genuinely thought that the company could die. And this was a really harsh moment, you know, when you're building a company and you've been toiling really hard at work for months, and you don't quite know if you're gonna make it. So January 2022 was: no money, no investors, no partners. Tough times.

Fast forward to 2023, we raised our first round of 1.5 million euros. It was announced in December last year, with a really really strong network of investors and partners. We got Elaia in France, Bessemer Venture Partners in the US, as well as a strong set of experts in AI, in business, in software, like the CTO of Hugging Face, the founder of the Good in Tech research foundation, the CEO of McCourt Global, the ex-CTO of Uber and really many more, that are not only investors but also really advisors to the company.

Also, I want to mention a new set of partners, really expanding our reach. First, partners related to AI standards. So we are working with ISO, with AFNOR, with CEN-CENELEC, so it's French, International, European levels on the standardization of AI. Moreover, we were selected last year in the FrenchTech DeepNum20 program. It's a program for the top 20 companies working on DeepTech innovation in France. We were really honored by that. Lastly, I want to announce, I'm really happy to say that we are a part of Cap Digital Paris region, a new group dedicated to responsible digital innovation. We're also in Systematic for DeepTech.

I want to give a big shout out to my co-founder Jean-Marie, who was the first on TV.

Our first interview on BFM TV Tech & Co

We literally, funny anecdote, we literally heard about this opportunity to speak about AI risks and security the day before. And Jean-Marie was instantly like, yes of course, I'm going to do it. We had never been on TV before. I think he did a fantastic job, so I encourage you to watch it, I put the link on the slides. It's really like explaining, first the problems: what are the risks, talking about Generative AI, the new wave of AI, including chatGPT, and how concretely Giskard can help, with our platform, with our tests and safeguards for AI models.

💝 Customers

Speaking about the customers, the heart of the company. Last year in January 2022, we really had a handful, so three early adopters who had signed up to our beta: Webedia, CrossData, TheFork, a small set of believers to whom we are super thankful.

But now in 2023, this community is really thriving. We got over 10 companies using Giskard, big names across company sizes, industries: Webedia in the media space, we got some startups that are pure players in AI, larger financial services firms, a whole panel of companies to whom AI quality, really, really matters. We also have a blooming developer community with nearly 1700 members, counting GitHub, Twitter, Discord.

So we're really seeing a rising interest around AI Ethics, Robustness, Standards. And this was really such a leap forward compared to when we had started to think about this idea of AI quality, several years ago. And this was not a concern, a lot of people took us for crazy persons to think that Machine Learning engineers and Data Scientists would care about the resilience, the robustness and the ethics of their systems. Today that topic is increasingly top of mind, and that's where I think lies the reason why the community for our product is developing.

💻 Product

Now let's talk about the product, the most important thing. In January 2022, we did have a product, which was a closed-source SaaS platform, allowing to explain AI models with human feedback. That feature still exists today, but it was fairly limited: the fact that it was closed-source, that it was a SaaS, was not really helping on the adoption side.

So fast forward to today, we are an open-source AI Quality Assurance platform, with both the collaborative AI inspection part, and also a very big feature which landed last year: Automated AI testing.

Giskard's Quality Assurance platform for AI models

We also plan, stay tuned, at some point to expand into AI Remedy: a way to fix AI biases and errors with synthetic data augmentation, so Generative AI helping to remove the biases of other AIs. This is packaged as open-source with free self-hosting, and we also have a paid professional offering where we add more security, integrations and support.

🎉 Special announcement

Big announcement today: I'm proud to say that our product got featured in the Gartner guide for AI Trust, Risk and Security Management. We are really really proud of being featured there, and hope it raises awareness especially to global organizations that really care about AI Trust, Risk and Security.

Giskard in the Gartner Market Guide for AI Trust, Risk and Security Management

⛵️ What's next? 2023 🐢🕊️

So what's next on the product level? Giskard 2.0.

The main big feature of Giskard 2.0 will be Data Slicing: a way to identify business slices where your AI models underperforms. Data scientists may have a good aggregate view of performance, but often, where AI incidents can happen, is in small slices of data that were left unchecked, kind of an inherent vice. This feature is going to help identify these small slices, that in production could lead to catastrophic AI performance issues. Of course, you'll be able to design these slices with business experts, to save them, to have a set of lenses from which you will evaluate your model performance.

Giskard 2.0 - Data Slicing

Next in Giskard 2.0, we'll offer AI Scan, a feature to detect the vulnerabilities and the biases of your AI model, right from your Python notebook. That feature will be instrumental to evangelize, to help the Data Science community know about the risks about AI, in a very actionable way, right in their development environment.

Giskard 2.0 - AI Scan

Lastly, our vision is to have one Quality Assurance platform to rule them all. We started with tabular AI models, classification and regression, also branched into NLP, with classification of NLP models. But we have them all in our list. We have already started to prototype how to test and inspect Computer Vision models, done the same for Generative NLP models, so Large Language Models. And next, we'll do the same for Recommender Systems and Time Series.

Giskard's vision: One platform to rule them all

Lastly, we're going global. We are a French company with big ambitions. First, we're doing a tour all around Europe. So this Saturday February 4th, we'll be at FOSDEM, one of the largest open-source conferences in the world. Next week, February 9 to 10, we will be at the World AI Cannes Festival in France. Then, in March, in London at the Big Data and AI World. We'll also be online, covering Sweden, with the Sweden Innovation Days, March 21st to 23rd. And we'll be in Frankfurt, Germany on May 10 to 11.

Giskard 2023 Conference Tour

We're also very interested to expand to other areas of the world, especially North America. So if you're interested to meet our team, invite us to an event or give a talk, give us a shout.

Right, that was it. I hope you found this new format, the video, interesting. Of course, there will be a transcript posted on our blog.

Thank you so much, and see you soon!

Integrate | Scan | Test | Automate

Giskard: Testing & evaluation framework for LLMs and AI models

Automatic LLM testing
Protect agaisnt AI risks
Evaluate RAG applications
Ensure compliance