Dear dida follower,
Happy New Year from the dida team! We would like to kick off 2025, with an early newsletter, partially looking back at our highlights from the second half of 2024, and also making a few announcements for the new year.
Let us directly start with the latter:
We are excited to announce that dida will organize its third dida conference on May 23rd, 2025 at frizzforum in Berlin. What started as an experiment in 2023, was so well received that we decided to continue with it. Due to the high demand from the previous years, we moved our conference to the new location, which will offer more space. Your feedback encouraged us to once again offer a mix of technical and applied presentations around AI and Machine Learning. In the past, we’ve had speakers and panelists from companies like Google, Bundesdruckerei, Idealo, Enpal or APCOA - and for 2025 we’re already busy putting together an equally impressive program. As the number of participants is limited, we recommend an early free registration here.
Looking back at some highlights from 2024, we’re very happy that dida has secured two large framework contracts by ITSV (IT-Services der Sozialversicherung GmbH). Over the next four years, dida will implement AI projects for Austria’s healthcare and social insurance providers, focusing on natural language and image processing projects. While for the natural language projects dida acts as the main contractor, we will collaborate with our trusted partner EBCONT GmbH for the image analysis projects.
All major social insurance institutions in Austria, especially from the healthcare sector, are entitled to request our services, including the Austrian Health Insurance Fund (“Österreichische Gesundheitskasse”) with 8 million insured members, the General Accident Insurance Institution (“Allgemeine Unfallversicherungsanstalt”) with 4.7 million insured members and the Pension Insurance Institution (“Pensionsversicherungsanstalt”) with 3.6 million insured members. For inquiries regarding social service projects, please reach out to Julius Lauenstein, our public services representative.
In addition to public sector projects, we would also like to introduce two additional new projects:
In terms of dida’s product developments, our information extraction software smartextract is making progress: With our first SaaS solution, we’re promising to automate data entry and validation processes with time savings of up to 95%. smartextract applies latest LLMs and does not require any training data or lengthy adaptations. It truly understands complex free texts in emails and PDFs - based on content, not patterns, and we’re glad to see customers retrieving value from it, such as Dennis Schöddert from Rudolf Peters Landhandel GmbH & Co. KG:
“We were able to reduce data processing from hours to minutes. This solution not only saves us time, but also numerous errors. We were particularly impressed by how quickly the software went live without a training phase.”
You can open a trial account and test it out for free within a few minutes. If you prefer a private demo, you can simply talk to our Product Lead Axel Besinger.
With respect to recent research publications, dida was able to place an article in the Journal of Machine Learning Research (JMLR) titled “From continuous-time formulations to discretization schemes: tensor trains and robust regression for BSDEs and parabolic PDEs.” The JMLR is one of the highest recognized journals in the field of machine learning and we’re very happy to be allowed to contribute to it, by addressing the numerical approximation of partial differential equations in high-dimensional contexts. We could furthermore present our research work about generative modeling at the Fourth Symposium on Machine Learning and Dynamical Systems in Toronto, Canada, at the prestigious Fields Institute for Research in Mathematical Sciences.
Finally, coming to an end, dida continues to focus on LLMs, with several projects currently in progress. After having mainly focussed solely on the implementation of open-source-based LLM projects, we’re now expanding our offering to the world of proprietary LLMs, also offering custom development services for technology stacks like OpenAI’s models through Microsoft Azure or through ChatGPT Enterprise. For more information about our OpenAI-based services, please contact David Berscheid. If you’re interested in more LLM content, we recommend our latest blog article about “Hallucinations in LLMs: Strategies for monitoring” by Şevval Gölbasi.
We once again say thank you for your continued interest in dida. If you have questions or would like to discuss any of the topics mentioned, please send us an email. We’re always happy to meet and exchange for a round of updates.
Philipp Jackmuth & Dr. Lorenz Richter
