I design and build full-stack frontend architectures for complex data, search, and AI-assisted applications.

My work sits where product UI, retrieval systems, and implementation detail meet: interfaces for large datasets, explainable search, semantic retrieval, structured knowledge platforms, and tools that make system behavior understandable to users.

• Full-stack Dev and frontend architecture with Vue, Nuxt, TypeScript, APIs, and data pipelines
• Requirements engineering for complex products: stakeholder goals, essential use cases, user stories, acceptance criteria, and prioritization
• Search UX and retrieval interfaces for complex domains
• Elasticsearch, semantic search, RAG, vector retrieval, and ranking workflows
• Data-heavy product UI: filtering, comparison, dashboards, metadata views, evidence panels
• AI-assisted workflows that expose sources, confidence, diagnostics, and user control

Research infrastructure for audiovisual metadata and film-related materials.

• Complex search and discovery across hierarchical metadata
• Vue/Nuxt frontend architecture for data-heavy research workflows
• Elasticsearch-backed search and structured result presentation
• Focus on usability, metadata clarity, and domain expert workflows

https://github.com/AV-EFI

Local, project-specific RAG system for developer and research workflows.

• Qdrant + Ollama retrieval pipeline with strict project separation
• Code-aware ranking, HyDE, source weighting, and retrieval-only modes
• MCP tools for IDE assistant workflows, feedback capture, and self-optimization
• Benchmark and real-question suites for retrieval quality and profile tuning

https://github.com/steffolino/local-rag

Knowledge platform for German social services, benefits, aid programs, organizations, and support contacts.

• LLM-ready structured data architecture
• Crawling, enrichment, moderation, and evidence-aware retrieval pipelines
• Official, assistive, and contact answer lanes for social-service guidance
• Easy German fallbacks, answer-shape guards, and retrieval diagnostics

https://github.com/steffolino/systemfehler

Open civic PWA for finding, reviewing, and reporting public toilets in German cities.

• Community-driven public-data product with multilingual UX
• Structured location data, user reports, and contribution workflow
• Product UI for maps, civic data, reviews, and accessibility-relevant information

https://github.com/steffolino/sandra-loo

Chrome and Firefox extension prototype for privacy-first comment-risk detection.

• Local-only rule-based analysis; no comment text leaves the browser
• Content scripts, popup UI, MV3 background worker, and platform adapters
• Inline overlays for visible comments without modifying original content
• Safety boundaries around suggestions, false positives, and user control

https://github.com/steffolino/trollguard-extension

Playable Nuxt/Vue survival roguelite card-game MVP used as an interaction-design experiment.

• Draft flow, rival groups, daily actions, d6 resolution, events, and recruitment
• Data-driven game rules with typed Vue/Nuxt UI components
• Visible probability, roll math, before/after state changes, and outcome feedback
• Useful testbed for complex state, progressive disclosure, and player-facing system clarity

https://github.com/steffolino/thegame

• Leipzig Open Data: 3D civic-data visualization with ArcGIS, NO2 sampler timelines, and map-driven exploration
• LLM UI Benchmark: controlled local prompt comparison for UI generation quality, accessibility, and instruction-following
• Knowledge graph editor: domain-expert editing UX for structured graph data without requiring RDF/SPARQL knowledge

I care about frontend architecture as a product capability: clear state models, resilient data loading, accessible controls, predictable navigation, and interfaces that help users understand why a system returned a result or made a recommendation.

Before implementation, I like to make the problem explicit:

• Identify stakeholders, user groups, constraints, and decision points
• Map essential use cases before jumping into screen-level UI
• Turn fuzzy needs into user stories, acceptance criteria, and testable workflows
• Separate business rules, user goals, edge cases, and implementation assumptions
• Prioritize with stakeholders so delivery decisions are visible instead of accidental

For AI and search systems, I usually work from the evidence outward:

• What sources exist?
• How are they ranked?
• What should the UI reveal?
• Where can users inspect, compare, correct, or reject the result?
• Which failure states need to be explicit?

That combination matters because complex systems rarely fail only in code. They also fail when requirements are ambiguous, stakeholders are misaligned, or the UI exposes implementation detail instead of user intent.

Portfolio: https://stefanstretz.de

Available for:

• Full-stack frontend architecture for complex data products
• Requirements engineering, essential use cases, and stakeholder workshop support
• Search UX and Elasticsearch-backed interfaces
• RAG, semantic search, and retrieval evaluation
• Vue/Nuxt product development
• Data-heavy dashboards, metadata tools, and AI-assisted workflows

Upwork: https://www.upwork.com/freelancers/~010d9cdbf11b00a5ca

Also open to direct contact.