Taking AI on sensitive data into production.

I’m Tim Fiedler, a computer scientist and AI enabler based near Oldenburg, Germany. I take AI on internal data into production for companies in finance, MedTech, industry and energy. The foundation is over 20 years of software and data engineering: architecture and production-ready code from one person.

Project experience with

  • dmTECH
  • DEUTZ AG
  • Volkswagen
  • Union Investment
  • Generali
  • Suva
  • EnBW
  • Engie Deutschland
  • RTL Broadcasting
  • NEXT Kraftwerke

0x01 Services

AI in production, built on a foundation of data and engineering.

AI enablement & agentic development

Generative AI should do productive work on internal data rather than stall as a pilot.

From use case to enterprise-wide rollout. RAG and LLM architectures on internal data, tool integration via MCP, introduction of AI-assisted development in the team with mandatory reviews and tests. Models from the cloud or, where data must not leave the company, with Ollama and LiteLLM on your own infrastructure; evaluation, requirements and operating costs decide.

Reference: DEUTZ AG, Union Investment, dmTECH

Data & Google Cloud

AI is only as good as the data it can access and the platform it runs on.

Data pipelines, data warehouses and data mesh architectures with BigQuery, Snowflake, dbt and Airflow. Platform operations on Google Cloud with infrastructure as code (Terraform/OpenTofu), Kubernetes and cost optimisation.

Reference: dmTECH, OVYAPP GmbH, RTL Broadcasting

Software engineering & architecture

Critical software needs to be modernised, scaled or put on a sustainable architecture.

System design, microservices and event-driven architectures, production-ready code in Python and Rust, modernisation of legacy systems. Test-driven development, code quality standards and CI/CD are an integral part.

Reference: NEXT Kraftwerke, RTL Broadcasting

0x02 Projects

Selected projects

A selection of my projects as architect, tech lead and engineer, from large enterprises to mid-sized companies.

DEUTZ AG

Mechanical engineering

Tech Lead (task force, part-time) & AI/RAG Architect

Enterprise-wide RAG rollout with Vespa and Onyx. Secured against prompt injection and data poisoning, strict IAM policies. Technical lead of the backend team.

  • Azure
  • Python
  • FastAPI
  • RAG
  • Vespa
  • Onyx
  • Grafana

Union Investment

Finance

Principal Data Engineer & AI Architect

Secure RAG architecture on internal bank data. Automated testing of LLM outputs. Hardened containers for high-security zones.

  • Azure OpenAI
  • LangChain
  • Vector DBs
  • Celery
  • RabbitMQ
  • Kubernetes
  • LLM Evaluation

dmTECH

Retail (IT)

GCP Cloud Architect & Data Engineer

Data pipelines with Airflow on GCP. Infrastructure as code, cloud cost optimisation. AI enablement of the team, including AI-assisted coding.

  • GCP
  • Python
  • Airflow
  • Terraform/OpenTofu
  • Cloud Cost Optimization
  • AI-assisted Coding

OVYAPP GmbH

MedTech

Tech Lead & Cloud Architect

MDR-compliant Google Cloud infrastructure for health data. Hardening and strict IAM policies. Technical lead of the backend team.

  • GCP
  • BigQuery
  • Python
  • Flask
  • MDR-Compliance

RTL Broadcasting

Media

Tech Lead Big Data

Big data platform on Google Cloud. Heterogeneous data streams unified. Rust and Polars for fast processing.

  • Rust
  • Polars
  • Python
  • GCP
  • BigQuery
  • Airflow
  • Kubernetes
  • Terraform

NEXT Kraftwerke

Energy trading

Senior Python Developer & DevOps

Refactoring and architecture of task queues in a critical infrastructure (KRITIS) environment. Migration to Celery and Redis for higher resilience. CI/CD with security scanning.

  • Python
  • Django
  • Celery
  • Redis
  • Docker
  • Kubernetes
  • GitHub Actions

Further projects

Engie Deutschland
Data mesh and data hub architecture in the energy sector.
EnBW
Data warehouse with Snowflake, dbt and Airflow.
re-minds
Invoice processing migrated from Python to Rust.
Earlier
EWE Trading, FITKO, MöbelFirst, Nachtblau, and web development with PHP and Symfony.

0x03 Testimonials

What clients say

His technical expertise is very high, and as a contractor he discusses the best way forward for us. […] Full recommendation, both professionally and personally.
MöbelFirst Management
He was able to take on the requirements and implement them quickly and professionally. Communication and technical expertise were flawless, which helped us greatly given the project’s tight deadline.
lifecon GmbH Sales
He stood out for his very strong expertise and exceptional commitment. Particularly noteworthy is his balance of a hands-on mentality with the aim of completing tasks thoroughly and sustainably.
HFR Senior Consultant

Excerpts translated from the German reviews on freelancermap.

0x04 How I work

Engineering discipline, including for AI

I deliver tested, secured systems with traceable decisions, and enable your team to run them on its own afterwards.

Security by design

Security is decided in the architecture, not added afterwards. Permissions, data flows and hardening are part of every design.

System design before tool choice

Requirements, data flows and risks come first, then models and services. Language model outputs are tested like any other code, and whether a cloud or a local model fits is an architecture decision.

Enablement, not dependency

Code quality standards, TDD, reviews and mentoring are part of it, including for AI-assisted development. Your team runs and develops the result on its own afterwards.

How we get started

  1. Enquiry

    A few sentences about your project and timeline are enough. I usually respond within 24 hours.

  2. First call

    We clarify goals, risks and constraints. If the project doesn't fit my profile, I'll say so directly.

  3. Proposal

    You receive a concrete proposal covering role, scope and start date.

0x05 Tech stack

Technologies used in projects

AI enablement
RAG, LLMs, LLM Evaluation, MCP, AI-assisted Coding, LangChain, Vespa, Qdrant, Onyx, Ollama, LiteLLM
Data
BigQuery, Airflow, dbt, Snowflake, Polars, Data Mesh
Google Cloud & platform
GCP, Kubernetes, Terraform/OpenTofu, GitHub Actions, Cloud Cost Optimization
Software engineering
Python, Rust, FastAPI, System Design, Event-driven (RabbitMQ, Kafka), TDD, Legacy Modernization

0x06 Blog

Latest articles

View all articles

0x07 FAQ

Frequently asked questions

Who is Tim Fiedler?

A freelance AI enabler and IT engineer based in Wiefelstede near Oldenburg, Germany. I take AI on sensitive company data into production, built on a foundation of software and data engineering. My cloud platform is Google Cloud; where data must not leave the company, models run on its own infrastructure. Over 20 years of experience in software development and architecture. Bachelor of Science in Computer Science from the University of Applied Sciences Emden, focus on IT security and media informatics.

What does AI enablement mean for you?

That AI doesn’t stall as a pilot. Together we clarify use cases, risks and compliance; I design the architecture, build it production-ready and enable your team to carry the solution forward on its own. At DEUTZ, for example, I had technical ownership of the RAG strategy, compliance and the enterprise-wide rollout.

How do you use agentic software development?

As a tool with clear rules, not as a substitute for engineering. I introduce AI-assisted and agentic coding in development teams, currently at dmTECH, for example. Generated code goes through the same reviews, tests and security checks as any other code. I connect tools and data via the Model Context Protocol (MCP) with restricted permissions; one example is the public, read-only MCP server of this website.

What experience do you have with RAG in production?

At DEUTZ, design and enterprise-wide rollout of a RAG application, secured against prompt injection and data poisoning. At Union Investment, design of a secure RAG architecture for language models on internal bank data and an automated test suite validating LLM outputs.

Can language models be run locally?

Yes, when confidentiality, data protection or independence from individual vendors require it. The models then run with Ollama on your own infrastructure, with LiteLLM providing a single interface to local and cloud models. I run a LiteLLM gateway with various models myself and use it daily. Whether a local model is sufficient or a cloud model is the better choice is decided on requirements, evaluation and operating costs.

Do you also offer IT security?

Yes, as a separate offering at ringnull.tech: AI security, IT security and IAM, social engineering and IT forensics. This site, 0x68.de, covers AI in production and the data and engineering foundation behind it. Security is part of the architecture there, not a separate package.

Which cloud do you work on?

I specialise in Google Cloud, for example at dmTECH, OVYAPP, RTL and MöbelFirst. I also have project experience on Azure (DEUTZ, Union Investment, Engie) and AWS (EnBW).

What are you not the right fit for?

Pure frontend, app or CMS projects. My field is AI, data and software engineering on Google Cloud.

Do you also work on site in Oldenburg and Ammerland?

Yes. I’m based in Wiefelstede in the Ammerland district, about 15 kilometres north of Oldenburg in Lower Saxony, Germany. For workshops, kick-offs and architecture reviews I’m happy to come to you within about 150 kilometres of Oldenburg, for example in Ammerland, Bremen, East Frisia, Emsland and much of Lower Saxony. Ongoing delivery is mostly remote, including for clients across Germany, Austria and Switzerland.

When are you available?

For a new project from August 2027, after my current engagement ends. Shorter engagements such as architecture reviews or workshops are possible sooner by arrangement. On site within about 150 km of Oldenburg, remotely across the DACH region.

How does an enquiry work?

Use the contact form on this page or email tim@0x68.de. I usually respond within 24 hours.

0x08 Contact

Tell me about your project.

I usually respond within 24 hours.

Next project can start from August 2027, reviews and workshops sooner

Prefer email? tim@0x68.de

Your details are sent to me by email and used only to process your enquiry. Read more in the privacy policy.