07.11.2026 aktualisiert

**** ******** ****
Premiumkunde
100 % verfügbar

Agentic Software Engineer / Machine Learning / AWS & Azure / Test Automation

Luzern, Schweiz
Luzern +100 km
PhD in Physics (University of Toronto)
Luzern, Schweiz
Luzern +100 km
PhD in Physics (University of Toronto)

Profilanlagen

CV-Wolfram-Helwig-EN.pdf
CV-Wolfram-Helwig-DE.pdf

Über mich

A Swiss Army knife — now actually based in Switzerland. Coding since childhood, physicist (PhD), came to algorithms through game theory and from there to cloud, machine learning and AI agents. Curious by nature, quick to find my way around new domains. Lately: software built with and for AI agents.

Skills

Künstliche IntelligenzApache AirflowMachine LearningTensorFlowAzure Machine LearningSqlalchemyAWS CDKTest-Driven DevelopmentData ScienceAzure Data FactoryPyTorchLarge Language ModelsDeep LearningCloud TechnologiesGitLabPytestData LakeScikit-learnTerraformJenkinsDatabricks
Freelance agentic software engineer — LLM applications and AI agents, machine learning and cloud (AWS, Azure). Physicist (PhD) with more than 10 years of experience in software development, data science and test automation.
Based in Lucerne, Switzerland — remote-first, on-site in the Lucerne / Zug / Zurich area by arrangement.
Core competencies:
  • Languages & frameworks: Python (PyTorch, TensorFlow, JAX, scikit-learn, pandas, NumPy, FastAPI, Django, Flask, SQLAlchemy, pydantic), Jupyter, TypeScript / JavaScript, React, Java, C++, SQL (PostgreSQL, DB2), GraphQL
  • Machine learning & AI: Deep learning (CNN, RNN, Transformers, attention), NLP & LLMs, generative models (diffusion, LoRA fine-tuning), reinforcement learning (DQN, PPO, A2C, MCTS, CFR, NFSP), classical ML (regression, random forests, gradient boosting, clustering)
  • AI agents: Agentic development with Claude Code and Codex: parallel agents in isolated working copies, independent review agents; LLM applications with tool integration over MCP (Model Context Protocol), separate permissions and resumable tasks; evaluation with deterministic checks and LLM judges
  • Cloud & data engineering: Azure (Machine Learning, Functions, Data Lake Storage, AKS, Databricks), AWS (Lambda, S3, DynamoDB, Kinesis, SQS/SNS, Step Functions, SageMaker, API Gateway), Docker, Kubernetes, Apache Spark / PySpark, Terraform, AWS CDK, Serverless Framework
  • Quality & testing: Test automation in large-scale programmes (Robot Framework, pytest), Robot Framework tooling (language server, debugger, test management), TDD, automated test and deployment pipelines, test-quality reporting, business process analysis (Camunda)
  • Ways of working: CI/CD (Jenkins, GitLab, Bitbucket), agile / Scrum, monitoring (CloudWatch, Sentry, Elastic / Kibana), infrastructure as code
Certifications:
  • Microsoft Certified: Azure Data Scientist Associate
  • Microsoft Certified: Azure Data Engineer Associate
  • Microsoft Certified: Azure AI Engineer Associate
  • Microsoft Certified: Azure Fundamentals
  • EXIN Agile Scrum Foundation

Erfahrung

Berufserfahrungmehr als 10 Jahre
Freelancer seit2021

Sprachen

DeutschMutterspracheEnglischverhandlungssicher

Projekthistorie

LLM Harness — agentic software analysis and development automation

Own project — built with AI coding agents
  • Designed and developed an application connecting language models to authorised repositories, documentation and tickets for source-backed analysis and controlled engineering workflows
  • Implemented agentic workflows from research and proposed changes to controlled file edits, with separate permissions for reading, writing, test execution and Git operations
  • Integrated structured Robot Framework execution and read-only database and process diagnostics for investigating test results
  • Built resumable tasks with saved progress, bounded retries and recorded failure and recovery events
  • Verified behaviour through automated Python and JavaScript tests, synthetic integration scenarios and independent reviews
Technologies used: Codex, Python, JavaScript, HTML/CSS, MCP, REST, Robot Framework, Git, SQL

Test Automation Framework — runcomposer and Test Manager

Own project — built with AI coding agents
  • runcomposer: API-first platform for orchestrating large automated test suites — tag-based selection with a compile step that previews exactly which cases a filter matches, run configuration and CI handoff; FastAPI backend with pydantic contracts, React/TypeScript front end; in use in a live test-automation mandate, MIT licence — github.com/StochasticEntropy/runcomposer
  • Test Manager (prototype, since 09/2026): business users with no knowledge of test code find, visually assemble and run automated tests in German or English business language, executed with Robot Framework; tested on generated repositories of up to 1319 tests
  • Developed agentically: coding agents implement tasks in isolated working copies; an independent review agent checks evidence and test results before a task counts as done
Technologies used: Python, FastAPI, pydantic, React, TypeScript, JavaScript, Robot Framework, pytest, Jenkins, YAML, Chromium (headless), Claude Code, Codex, Git

Senior Software Engineer

Insurance
  • Designed a green-field software architecture in Python for a large-scale test automation project
  • Developed the Python keyword libraries and the business test logic behind the Robot Framework suites
  • Implemented the backend data access layer (SQLAlchemy, DB2) and the REST/SOAP integration of further components
  • Built and operated the pipelines for automated tests and deployments (Jenkins, Ansible)
  • Developed an internal reporting tool (FastAPI) for the daily regression overview — failure patterns, annotations and history for the team
  • Analysed business processes in Camunda as the basis for test cases
  • Debugged the Java application layer (IntelliJ)
  • Agile collaboration in a distributed team (Scrum, Jira, Confluence, Bitbucket)
Technologies used: Python, Robot Framework, FastAPI, SQLAlchemy, SQL, DB2, Jenkins, Ansible, REST, SOAP, pydantic, Java, Camunda, Bitbucket, Git, Jira, Confluence

robot-companion — Robot Framework environment for VS Code

Own project — built with AI coding agents
  • Editor tooling that surfaces return values, enums and types for Robot Framework: documentation preview and return explorer in the sidebar, hover information and type-matched autocomplete from the indexed Robot and Python documentation
  • Own Python backend for the full Robot Framework workflow in VS Code: language server with navigation, diagnostics, completion, references, rename and formatting, test discovery in the Test Explorer, and a debugger for mixed Robot/Python code — no other Robot Framework or Python debug extension needed
  • Built with AI coding agents; published on the VS Code Marketplace under the MIT licence, the standalone backend as a preview release — github.com/StochasticEntropy/robot-companion
Technologies used: JavaScript, Node, VS Code Extension API, Python, Language Server Protocol, Debug Adapter Protocol, debugpy, Robot Framework, Claude Code, Git

Comic Workshop — agentic production line for a 114-page illustrated book

Own project — built with AI coding agents
  • Claude agents and an image model produced a 114-page comic and a 19-page making-of through to the print PDF: one agent strand per location, up to 16 pages in parallel; one source for every fact in YAML and a deterministic prompt builder with no AI in it
  • Quality assurance as a mix of deterministic and agentic checks: scripts verify text, reading order and panel layout; agents check cast, states and every page panel by panel; a reader who never saw the story reconstructed it from the prompts alone (957 of 978 story points)
  • Judge rules from production: no judge approves a repair it triggered, a missing answer counts as a failed check, new judges are first tried on known errors — guide: claude.ai/artifact/4g1o82myb8SWk5RWM4oq7N
Technologies used: Claude Code (Opus, Sonnet), Codex with OpenAI image model, Python, OpenCV, Apple text recognition, YAML, Real-ESRGAN, print PDF production

Studio for generative narrated videos

Own project — built with AI coding agents
  • Fully local studio on Apple Silicon for AI-generated narrated videos — images, video, narration and editing on the machine, with automatic quality checks; designed and built largely by AI coding agents
Technologies used: Claude Code, Codex, MCP, Python, Remotion, SDXL, FLUX, LoRA, Wan 2.2, Qwen3-TTS, Whisper, ffmpeg, MLX, Apple Silicon

mnemo — memory engine for AI coding agents

Own project — built with AI coding agents
  • Tiered short-, mid- and long-term memory across agent sessions; autonomous nightly consolidation runs with auditable reports and enforcement gates — github.com/StochasticEntropy/mnemo
Technologies used: Claude Code, Codex, agent skills, MCP, Git, launchd, shell/zsh, Python

Senior Backend Engineer

Smart mobility
  • Implementing AWS Lambda Serverless code to process incoming Kinesis streaming data and Frontend requests.
  • Frontend development with JavaScript and React
  • Containerized deployment of backend code with Docker and FastAPI
  • Bug tracking with Elastic, Kibana and CloudWatch
  • IaC with Terraform and Serverless Framework
Technologies used: AWS Lambda, API-Gateway, SQS, SNS, Kinesis, Terragrunt/Terraform, Serverless, Elastic, CloudWatch, Python, Git, PostgreSQL, Hasura, GraphQL, REST, Docker, GitLab

Python Cloud Developer

E-commerce
  • Developed a serverless, API-driven data management system on AWS (Lambda, DynamoDB, SNS, SQS, API Gateway)
  • Implemented the AWS Lambda layer between customer API requests and the backend NoSQL DynamoDB and S3 data stores
  • Developed a BERT-based NLP model and integrated it into the serverless architecture
  • Maintained the OpenAPI specifications; monitoring via CloudWatch and Sentry; IaC with AWS CDK
Technologies used: AWS Lambda, DynamoDB, S3, Step Functions, API Gateway, SQS, SNS, AWS CDK, Terraform/Terragrunt, CloudWatch, Sentry, Python, Hugging Face, Rockset, SQL, Bitbucket, Git

Coach for Machine Learning / Deep Learning

Automotive
  • Delivered an interactive course to a group of 20 participants
  • Presented classical machine learning tools and concepts based on Python, numpy, pandas and scikit-learn
  • Presented deep learning tools and concepts based on TensorFlow v2, including CNNs, RNNs and Transformers
Technologies used: Python, TensorFlow v2, Keras, scikit-learn, numpy, pandas, Jupyter Notebooks, Azure VMs

AI & Cloud Engineering

Own projects: game theory & deep reinforcement learning
  • Developed and applied deep-learning methods: computer vision (CNNs, vision transformers), NLP (transformer-based models including BERT)
  • Deep reinforcement learning for imperfect-information scenarios (DQN, REINFORCE, A2C, PPO) and game-theoretic methods (MCTS, CFR, NFSP)
  • Designed and backtested quantitative trading algorithms with systematic result analysis
  • Implemented and reproduced recent research results in deep and reinforcement learning
  • Data science and distributed machine learning on Microsoft Azure and Databricks — the groundwork for the subsequent Azure certifications
Technologies used: Python, TensorFlow, Keras, PyTorch, JAX, Hugging Face, scikit-learn, NumPy, pandas, Jupyter, Docker, Apache Spark, Azure ML, Databricks, Data Lake Storage, GCP, TPUs

Algorithm Developer

  • Developing algorithms for the simulation of milling processes based on 3D-CAD data
  • Geometric complexity analysis of the milling process based on volume and surface properties
  • Implementation of the developed methods (Java)
  • Test-driven development (JUnit)

IT Architecture Consultant

  • Consulting in the exploration phase of a new Big Data System
  • Collection of business and technical requirements for the new system
  • Derived and presented suitable IT-architectures based on the collected business and technical feedback

Algorithm Developer

  • Mathematical optimization of algorithms for distance field calculations in a 3d analysis software
  • Implementation of the developed methods (Java) from prototyping to release.
  • Test-driven development (JUnit)

IT Consultant

  • Quality assurance of in-car control unit tests
  • Developed tools (Java, Python) for automatic test quality reporting

IT Consultant

  • Design and specification of a customer-specific software application to be in accordance with WLTP standards
  • Development of the required data structure
  • working in an agile environment

Quantitative Strategies

Own project: game theory
  • Built abstractions for computationally unsolvable strategic scenarios under imperfect information
  • Game-theoretical analysis: derived optimal strategies (Nash equilibria) over those abstractions
  • Generalised the resulting insights back to the full problem
  • Applied machine-learning methods (decision trees, clustering) to identify strong strategies in collected data
  • Successfully applied the developed strategies under real-time conditions
Technologies used: C++, Python, scikit-learn, Jupyter, PostgreSQL, Visual Studio

Research / Teaching Assistant

  • Analyzed transformation by local operations of multi-party distributed entangled quantum states in high dimensional Hilbert spaces
  • Theoretical study of the utilization of maximally entangled quantum states in cryptographic protocols and error correction algorithms
  • Proved conditions for the existence of maximally entangled state in a high dimensional multi-party settings
  • Computational investigation (Java) of subspaces of Hilbert spaces to find absolutely maximally entangled states
  • Lead tutorials for undergraduate students

Research / Teaching Assistant

  • Numerical and theoretical evaluation of multi-mode effects on the security of quantum cryptographic protocols
  • Wrote simulations in C++ to evaluate the security of quantum cryptographic protocols
  • Lead tutorials for courses in theoretical physics

Kontaktanfrage

Einloggen & anfragen.

Das Kontaktformular ist nur für eingeloggte Nutzer verfügbar.

RegistrierenAnmelden