Software Engineer · AI & Full-Stack Development
MSc Computer Science, University of Zurich (summa cum laude) · Switzerland
Tip
Open to software engineering roles from February 2027 — especially where I can ship AI products end to end in technology, finance, banking or consulting. Reach me at daniel.maksimovic01@gmail.com.
I design and build secure, reliable AI-enabled software end to end: from requirements and process modeling to back-end services, RAG and agentic LLM systems, and React front ends. Hands-on experience since 2019 across industry roles, internships, teaching and research.
- 🎓 MSc Computer Science, University of Zurich — graduated 2026, summa cum laude
- 🤖 Master's thesis at Ergon Informatik — a multi-agent AI teammate for software engineering and DevOps, evaluated with nine developers
- 🏢 Two years at Eraneos — developer on SafeGPT, Eraneos' privacy-compliant AI platform, and builder of a RAG advisory chatbot from concept to MVP
- 🧑🏫 Right now — Swiss civilian service (Zivildienst) as a teaching assistant, until February 2027
Master's thesis · University of Zurich × Ergon Informatik AG · Feb – Jul 2026 · Grade 5.75/6.0
An AI system that works as a shared member of a software team rather than one developer's personal assistant. An orchestrator coordinates specialized agents for coding, testing, pull-request review and log analysis, connected to Jira and Bitbucket. Together with Ergon's engineers I gathered requirements, integrated the system into a sandbox environment and evaluated it over three weeks with nine developers, measuring its impact on developer productivity and human–AI collaboration.
Stack: Agentic AI · LLMs · multi-agent orchestration · Jira · Bitbucket
Developer in an international Scrum team · Eraneos Switzerland AG
Eraneos' internal, privacy-compliant AI assistant platform: a secure multi-model chat solution that runs LLMs from several providers (e.g. Google Gemini and Anthropic Claude) inside a controlled environment. I contributed features, bug fixes and security improvements across the back end and front end, with code reviews in GitLab:
- RAG pipeline: token-based chunking of spreadsheet documents
- Security: server-side authorization checks for shared prompt collections
- Reliability: error handling for streamed responses
- Adoption: presentations, live demos, use cases with sales and back office, and a modular training program
Stack: Python · FastAPI · PostgreSQL + pgvector · React · TypeScript · Vite · GitLab
Master's project · University of Zurich · piloted for two weeks with licensed therapists
A therapist-centric no-code platform to design, supervise and evaluate AI counseling agents. Therapists create counseling plans and reusable exercises, configure client-facing chatbots (including document uploads), assign GAD-7 questionnaires, and review journals and transcribed meetings with AI summaries. A therapist assistant ("Eliza"), interaction monitoring with harmful-content alerts and client-activity dashboards support supervision. Built with a Design Thinking (Double Diamond) and agile process.
Stack: Java 21 · Spring Boot · PostgreSQL · Hibernate · TypeScript · React · Vite · MUI · OpenAPI Generator · Docker · Kubernetes · GitHub Actions · SonarCloud
LLMs: Gemma 3 12B self-hosted at UZH and Azure OpenAI GPT-4.1 (Sweden Central)
Built end to end, from concept to MVP · Eraneos Switzerland AG
An AI advisory chatbot for optimizing business networks, with RAG-based document retrieval, conversation memory and context-driven dialogue logic. It runs on locally hosted Llama models (AnythingLLM, LM Studio) and uses SCION for secure data transmission. Presented in live demos to Eraneos' TMT community.
Stack: React · Vite · Java · Spring Boot · Llama · RAG · SCION · Azure
| Project | What it is | Stack |
|---|---|---|
| No-code AI chatbot builder BSc thesis · Grade 5.5/6.0 |
A click-and-drop environment that lets non-technical users build educational or therapeutic chatbots and embed them in web applications. | React · Spring Boot · Node.js · GCP · SonarCloud |
| License server for Consor Universal Internship · Consor AG |
Automated license generation and delivery for Consor Universal, used by major insurers in Switzerland and Germany, with encrypted license access. | Java · Spring Boot · Docker · Jenkins · Kubernetes |
| Swiss equity token Blockchain Presence |
Extended an open-source ERC-20 contract for tokenizing Swiss shares along the CMTA blueprint, with a comprehensive Hardhat test suite. | Solidity · Hardhat · JavaScript |
| Taboo Online UZH Software Praktikum · team of 4 · client · server |
Real-time multiplayer Taboo with lobbies, team chat, a buzzer, invite links and QR codes. | React · Spring Boot · WebSockets · JWT · JUnit · Cypress |
- Teaching Assistant (Swiss civilian service) · Schule Pfaffnau · Aug 2026 – present
Supporting teachers and pupils in daily lessons, lesson preparation and classroom ICT, until February 2027 - Software Engineer Working Student · Eraneos Switzerland AG · Sep 2024 – Sep 2026
SafeGPT and the Business Network Advisor; evaluated Azure deployment options (App Service, AKS, AI Foundry, AI Search, Entra ID); drove internal adoption of SafeGPT - Software Engineer, Master's Thesis · Ergon Informatik AG · Feb – Jul 2026
Agentic AI system for software engineering and DevOps workflows - Teaching Assistant, Business Informatics II · University of Zurich · Aug 2025 – Jan 2026
Tutorials for ~80 students; designed and graded BPMN process-modeling exercises; exam preparation - Teacher, Computer Science & AI · Logiscool Switzerland · Aug 2024
CS and AI classes for ~30 students, including machine learning and ethical AI use - Software Engineer Intern (Java) · Consor AG · Jul – Sep 2023
License server for Consor Universal with a Spring Boot, Docker and Jenkins CI/CD pipeline - Computer Science Intern · FHNW · Jul 2019
Sensor-based Android app; ported Python logic to Android
- Joining Forces for AI-based Customer Feedback Management · Information Systems Seminar Conference, UZH, 2025 · Grade 6.0/6.0
Benchmarked 20 AI-enabled customer-feedback-management platforms by AI-role breadth and human–AI collaboration intensity, then combined a literature review with CEO interviews into a governance model for human–AI co-creation. - Hybrid Human–AI Teams in Digital Leadership: A Comprehensive Literature Review · University of Zurich, 2025 · Grade 6.0/6.0
Surveyed 80+ peer-reviewed studies to map challenges such as trust, communication and role ambiguity, and proposed ethical-governance and algorithmic-leadership guidelines.
- MSc Computer Science, University of Zurich · 2024 – 2026
Major in Information Systems, minor in Informatics · final GPA 5.5/6.0, summa cum laude
Thesis: Development and Integration of an Agentic AI System in Software Engineering (with Ergon Informatik, 5.75/6.0) - BSc Computer Science, University of Zurich · 2021 – 2024
Major in Information Systems, minor in Banking & Finance
Thesis: Development of a No-Code/Click-and-Drop Environment for the Development of AI-Chatbots (5.5/6.0)
Certificates: SAP Business ByDesign · Adcubum Syrius · Celonis Rising Star – Automation · Microsoft Startup Challenge
All grades use the Swiss scale, where 6.0 is the highest.
| Area | Tools |
|---|---|
| Languages | Java · Python · TypeScript · JavaScript · SQL · C · C++ · C# · Solidity |
| Back end | Spring Boot · FastAPI · Node.js · PostgreSQL · pgvector · Hibernate |
| Front end | React · Vite · MUI · HTML/CSS |
| AI | LLMs · RAG · AI agents · Azure OpenAI · local models (Llama, Gemma) · PyTorch · TensorFlow |
| Cloud & DevOps | Azure · GCP · Docker · Kubernetes · GitHub Actions · Jenkins · Gradle · Git · CI/CD |
| Quality & security | JUnit · Mockito · Jest · Cypress · SonarCloud · authentication & authorization · encryption · SCION |
| Process & enterprise | Scrum · SAFe · BPMN · process mining (Celonis) · SAP · Microsoft Dynamics · ERPsim |
🗣️ German (native) · English (professional) · French (good)
The fastest way to reach me is daniel.maksimovic01@gmail.com. More about me on danielmaksimovic.com.


