dida
A Berlin machine learning R&D firm, mostly computer vision, with one LLM agent build
- HQ
- Berlin, Germany
- Founded
- 2018
- Team
- 11-50
- Regions
- Europe
- Pricing
- From $25k Directory figures only: Clutch shows a $25,000+ minimum project and $150-$199 an hour, with typical budgets of €25,000-€50,000. Publishes no prices on its own site.
- Stack
-
- Python
- FastAPI
- LangGraph
- LangChain
- Docker
- AWS
- Snowflake
- Langfuse
- Salesforce
- Industries
-
- Ecommerce
- Real Estate
About dida
dida (dida Datenschmiede GmbH) is a machine learning research and development firm in Berlin, founded in 2018 according to Clutch and its footer. Its about page profiles 31 people, many with physics or mathematics PhDs, and names Dr. Lorenz Richter as CTO and co-founder. It sells custom ML builds, ML consulting and workshops, MLOps and research, and publicly funded projects for BMBF, ESA and Cyberagentur make up a visible share of the portfolio. LLM and agent work is the newer and smaller part of the business.
It publishes 23 projects, most with named clients: computer vision and remote sensing for Deutsche Bahn, ESA, DWD, APCOA and Enpal, pre-LLM text extraction for idealo, Felmo and Mieterengel, and one multi-agent LLM system for nVent's industrial customer service, built on LangGraph, LangChain, FastAPI, AWS Fargate, Snowflake and Langfuse. Clutch shows 5.0 from 6 reviews and Google 4.9 from 39, the Google ones unverified and mostly star-only. Clutch lists a $25,000 minimum and $150-$199 an hour.
No project reports a measured outcome, and the nVent figures are targets and design properties for a system still in pilot. The newest Clutch review is from February 2023 and all six describe classic ML, CV and NLP work from before the LLM practice. There is no G2, Trustpilot or DesignRush profile, no vendor partner tier or security certification, and the award claims on its Clutch profile are not documented on its own site.
Ratings elsewhere
Our verdict
dida suits a European company with a difficult computer vision, remote sensing or document ML problem that wants a research-trained team and has at least $25,000 to spend.
The case for it is depth and named work: 23 published projects with clients such as Deutsche Bahn, ESA and idealo, a team heavy with PhDs that publishes research, and an agent build for nVent whose architecture is described down to the stack.
The caution is that its agent record is one pilot. That project reports targets with no measured result yet, the outside reviews are from 2022 and 2023 and describe older ML work, and no project anywhere on the site states an outcome.
A buyer who wants a quick business workflow agent at a modest budget should look elsewhere; dida's pattern is research-grade custom ML.
Pros
- 23 published projects, most with named clients including Deutsche Bahn, ESA, DWD, idealo and APCOA
- One documented multi-agent build for nVent, naming its stack: LangGraph, LangChain, FastAPI, AWS Fargate, Snowflake and Langfuse
- Clutch 5.0 from 6 reviews and Google 4.9 from 39
- Research-trained team: the about page profiles 31 people, many with physics or mathematics PhDs, and the firm publishes papers
Cons
- Only one LLM agent project is published, it is still in pilot, and its headline 95% figure is a target with no measured result yet.
- The newest Clutch review is from February 2023, and all six describe classic ML builds that predate its LLM work.
- No project on the site reports a measured outcome.
- Publishes no prices; Clutch shows a $25,000 minimum and $150-$199 an hour, at the top of the European range.
- Much of the portfolio is publicly funded research and computer vision, so a quick business workflow agent sits outside its usual pattern.
Where dida ranks
-
Ranked #14 in AI Consulting Firms New in this ranking - Ranked #18 in AI Agent Development New in this ranking
Top three by industry
Each badge links to the ranking it comes from and can be embedded from the badges page.
Is this your agency? Get your badge.