Giskard

Giskard

Artificial Intelligence / Machine Learning, Software

Paris

Organization and methodology

We rely on the Kanban methodology for product development. Tasks usually vary from full-stack development to machine-learning model integration. The development team is composed of R&D engineers and ML researchers. Our roadmap is heavily influenced by the demands of our customers. When each project emerges, we organize brainstorming that every developer can contribute to.

Product, project or technical challenge

Check the quality of AI models for Computer Vision

Check the quality of AI models for Computer Vision

CV are one of the most widely used high-risk AI. For instance, it’s used for diagnosis using medical images to help doctors identify diseases at an early stage. Given the criticality, these models require special attention to bias & error mitigation by using highly-reliable quality testing. This projects aims to build the integration of CV models so that Giskard is able to inspect & test them. Three particular CV model types will be integrated : image classification, object detection and image segmentation.

This will require building ergonomic interfaces to inspect complex images and adapt Giskard test suites.

Check the biases of large NLP generation models

Over the last 2 years, the biggest AI breakthrough has been Large Language Models (LLMs) such as ChatGPT and Mistral. Despite their performance, these models raise many robustness and ethical challenges.

This project aims to integrate LLMs so that Giskard is able to inspect & test them. This requires computing metrics to measure distances between the predicted text and the right text output, custom filtering process to select examples and solutions to compute explanations of predictions.

This also requires building interfaces that enable ML engineers to design prompts, display text outputs and provide feedback on specific words.

Check the biases of large NLP generation models

Recruitment process

You can get an offer with salary & equity in 3 weeks 🚀

  • Fit interview: 15 minutes
  • Tech exercise: 10 days to complete
  • Tech interview: 45 minutes
  • Reference calls: 2 persons
  • Final interview: 45 minutes