Our Approach

Making Artificial Intelligence intelligible to businesses
Taking cutting edge Data Science beyond books, beyond labs, and into enterprises, for it to reveal hidden business value

Make it EMPIRICAL, RELIABLE, USABLE

Today, great research on AI is everywhere. The field now turns to experimentation that can be implemented quickly. The prime goal is to create value, fast. So we build state-of-the-art AI systems that work beyond the scientific papers.

 

Make it COLLECTIVE, COOPERATIVE, SHARED

The AI field is exploding in scope and complexity. It has moved beyond the reach of any single person. Solutions only come through teamwork by the specialists, who have deep expertise in their different domains. So we show users how to keep AI responsive through up-to-date knowledge and diversified teams.

people

LEARN FAST, LEARN WELL, LEARN ALL – WITH OPEN APPROACH

We bring to our clients a unique kind of value, not found in published scientific research. It works on several levels and leaves a unique mark. The cornerstone of our approach is openness. We keep no secrets when we work and we actively create opportunities for our users to understand how our solutions work in practice. This transfer of knowledge adds value through trust and a growing confidence, and builds the capacity for AI work in our clients’ teams.

comtrade-ai-approach1

TRUST MODELS,
TRUST PROCESS,
TRUST DECISIONS
– WITH EXPLAINABLE ALGOS

We build models that are transparent, helping users understand how they work. We show how AI makes decisions and suggestions, on two levels. We ensure that users can understand how a model decides, both in general and in each specific case. We find creative ways to achieve a high level of interpretability and explainability, qualities that make AI accessible to users. Our goal at every turn – explainable AI – the one characteristic of AI system that makes all the difference.

comtrade-ai-approach3

DECIDE FAST, DECIDE OFTEN, DECIDE WELL – WITH AGILE

On AI side, we collect and process data, label it, develop algorithms, train and evaluate models, then release a safe solution. Our software teams go through similar steps: they design, develop, test and deploy workable products in the meantime. The trick is to synchronize these two disparate tracks. We do it with passion, strong customer focus and a habit to share. Plus joint storyboards. The agile way.