Building AI prototypes
We work together with your team to identify where AI can be implemented. And then we prototype and test it with you.
Our methodology
The problem with building an AI product for everybody is that it will truly work for nobody. So we found a solution to find good problems: we call it the AI experiment
1. Training
1. Training
2. Research
2. Research
3. Prototyping
3. Prototyping
4. Evaluation
4. Evaluation
Training
Everything starts with a presentation by our researchers on the potential and limits of AI. At the end, we identify potential intrapreneurs in your team.
Research
We talk to your team in 1 to 1 interviews and identify challenges. Together with you we then select a few to work on.
Prototyping
Our AI engineers then develop prototypes for the challenges identified. These aim to test the performance of AI in the real world on your real tasks.
Evaluation
After a few months of your team using the prototype, we sit down with you to agree on next steps: improvements or deployments.
Training
Everything starts with a presentation by our researchers on the potential and limits of AI. At the end, we identify potential intrapreneurs in your team.
Research
We talk to your team in 1 to 1 interviews and identify challenges. Together with you we then select a few to work on.
Prototyping
Our AI engineers then develop prototypes for the challenges identified. These aim to test the performance of AI in the real world on your real tasks.
Evaluation
After a few months of your team using the prototype, we sit down with you to agree on next steps: improvements or deployments.
Training
Everything starts with a presentation by our researchers on the potential and limits of AI. At the end, we identify potential intrapreneurs in your team.
Research
We talk to your team in 1 to 1 interviews and identify challenges. Together with you we then select a few to work on.
Prototyping
Our AI engineers then develop prototypes for the challenges identified. These aim to test the performance of AI in the real world on your real tasks.
Evaluation
After a few months of your team using the prototype, we sit down with you to agree on next steps: improvements or deployments.
Technologies
AI is not just ChatGPT. As a team working with AI for more than a decade, we find the best tools to fit the problems,not the other way arround.
Classification
Text to voice
Transcription
Stable diffusion
Diarization
Constrained decoding
Classification
Text to voice
Transcription
Stable diffusion
Diarization
Constrained decoding
Classification
Text to voice
Transcription
Stable diffusion
Diarization
Constrained decoding
Translation
Fine tuning
Time series forecasting
Retrieval augmented generation
Image to text
Retrieval augmented generation
Translation
Fine tuning
Time series forecasting
Retrieval augmented generation
Image to text
Retrieval augmented generation
Translation
Fine tuning
Time series forecasting
Retrieval augmented generation
Image to text
Retrieval augmented generation
Video generation
Quantization
Model compression
Table question anwering
Sentence similarity
Prunning
Named Entity Recognition
Video generation
Quantization
Model compression
Table question anwering
Sentence similarity
Prunning
Named Entity Recognition
Video generation
Quantization
Model compression
Table question anwering
Sentence similarity
Prunning
Named Entity Recognition
Frequently Asked Questions
What do I get at the end of this “experimentation”?
Why to do this instead of a traditional procurement ?
What if you find no challenges at my organisation?
What if I don’t want to work like that ?
Can the public sector use this kind of process?
What do I get at the end of this “experimentation”?
Why to do this instead of a traditional procurement ?
What if you find no challenges at my organisation?
What if I don’t want to work like that ?
Can the public sector use this kind of process?
What do I get at the end of this “experimentation”?
Why to do this instead of a traditional procurement ?
What if you find no challenges at my organisation?
What if I don’t want to work like that ?
Can the public sector use this kind of process?