Experienced experts for AI and software development. Working directly in your project.
Deep Data Ocean is an owner-managed company of software developers. Our experts work in your project: in AI engineering, AI platforms and software development with C#, Java, Angular and React. What we know about AI in production we did not read, we built: in client projects and on trAIder, our own AI system.
How we work
You are looking for a specific person with specific skills for your project, not a company with slides. That is how we work: one of our experts works in your project for as long as you need them, whether you are short of capacity or of a particular skill. On this site you will find a complete profile for each of them, with projects, technologies and availability.
We come from project work ourselves. Founder Stefan Anders has worked as a freelancer in client projects for many years and knows every colleague’s work personally. Whoever contacts us talks to someone who understands the technology and gets an honest answer on whether the profile fits the task.
What our experts do
AI engineering
In your project we build systems with large language models, reinforcement learning and classical machine learning in Python and PyTorch. That includes the unglamorous work the result depends on: cleaning data, building features, making training runs traceable, and setting up backtests so that no information from the future leaks in.
AI platforms and operations
Models need an environment they run in reliably. We build GPU-capable Kubernetes platforms, run LLM inference with vLLM and deploy via GitOps with ArgoCD and Helm. Behind that are many years of platform work with OpenShift and Kubernetes, including in the public sector and in environments without internet access.
Software development with C# and Angular
For business applications on the Microsoft stack, an experienced developer of ours works in your project: a senior full-stack developer with almost 20 years of C# experience, from .NET Framework 4.8 to .NET 8 and ASP.NET 10, and with Angular frontends from version 1 to 21, including five years of development and operations on Azure. Sector focus: medicine and laboratory diagnostics, logistics and port operations, construction and explosive ordnance disposal, as well as HR and reporting applications for mid-sized companies.
Software development with Java and React
For business applications on the JVM, a Java developer of ours works in your project: backend with Spring Boot, REST interfaces and database access with JPA, frontends with React and TypeScript, delivered as containers on Kubernetes. The experience behind it ranges from large data platforms with Hadoop and Spark to breaking up Java monoliths into distributed architectures on AWS and Kubernetes.
Case study: trAIder
Our most demanding client is ourselves. trAIder is an AI-driven trading system for US equities that we have been developing and testing since 2025. Every trading day it evaluates thousands of stocks, makes its own decisions, and currently runs in forward paper trading; the first pilot with real money is tied to criteria set in advance.
On the way there we built a news pipeline that turned 2.65 million financial articles into features for reinforcement learning with Llama and Qwen models on vLLM, trained a Double DQN agent in PyTorch, and measured, discarded and documented both. What runs today is simpler and more rigorously tested: eleven million daily prices with verifiable provenance, walk-forward tests with random controls, guards against information from the future, a broker connection that does nothing when in doubt, and operations on Kubernetes that test their own alerts.
trAIder is in development and testing and is currently not offered to third parties. It shows how we work when we are the client.
Track record
The experience behind Deep Data Ocean comes from projects founder Stefan Anders worked in himself. It can be told in three stages that together explain why AI, for us, does not stop at the model.
It began with data. As Head of Development for Central, Eastern and Southern Europe at an international market research company, Stefan Anders was responsible for software development in 35 countries, with his own staff in ten of them, and replaced classic data warehouses there with Hadoop and Spark; data lake and streaming platforms for the automotive industry followed.
Then came the platform. Over several years he built and operated CI/CD and container platforms on OpenShift and Kubernetes as a DevOps architect, including for public-sector clients and for a cloud provider with installations fully disconnected from the internet. Added to that were migrations of Java and C# monoliths into distributed architectures on AWS and Kubernetes.
AI has run through this work since 2018: image recognition for visitor analytics in retail, text recognition for handwritten prescriptions, a neural network that automatically matches project requests to suitable skill profiles, LLM-based classification of financial transactions and, finally, trAIder.
Do you have a project to staff?
Send us the requirements. You will get a clear answer: who of us fits, and from when.
