Adrian Calma

Profil • Selbstdarstellung

I’m an experienced Machine Learning Engineer with a focus on practical solutions. I bring expertise in Human-in-the-Loop approaches and data quality to my work, leveraging established technologies. I rapidly build prototypes based on smart assumptions and iterate to refine them, favoring a practical and efficient approach over striving to outperform the state of the art.

Beyond technical expertise, I think strategically and work effectively with diverse stakeholders. I understand the importance of collaboration and communication, communicating technical concepts to non-technical stakeholders. My strength in working well with others enables me to deliver high-quality solutions that meet business needs.

With over 12 years of experience in data science and machine learning, I successfully co-organized 5 international workshops on Interactive Adaptive Learning, 4 at the European Conference on Machine Learning and 1 at the IEEE World Congress on Computational Intelligence. Since 2021, he is a member of the steering committee of the previously mentioned workshop.

My (industry) experience includes MLOps, fintech, annotation services, human-in-the-loop, loss prevention, computer vision for retail – brick and mortar–. As a German Accelerator in Silicon Valley alumni, I am successfully mentoring early stage, pre-seed startups like iurCrowd.de and myLawyer.ro regarding business model validation, technical guidance, and go-to-market strategies.

As a Professional Scrum Product Owner (>6 years), I have conducted over 300 customer interviews in the past 3 years. The interviewees had various backgrounds, from farmers and cashiers to senior machine learning engineers and CEOs. The customer insights helped the teams to address the user’s needs accurately and maximize their efficiency. For example, it took Signatrix just 2 weeks to build the MVP for the Bake-Off product and 4 weeks to the fully functional Demo with the customer’s representative.

At DB Systel GmbH, I led a team from inception to completion, enabling us to deploy Kubeflow in half the expected time.