Text-to-Speech Experts in Germany

Evidence-led health intelligence platform turning published research into personal health protocols. It scores 430 habits, foods, and supplements against the studies behind them, and moves the score when the evidence moves. Built solo.

• Built the daily ingestion pipeline across PubMed, bioRxiv, and medRxiv: 43,000+ papers from 3,400+ journals processed into 230,000+ typed evidence claims, each one traceable back to the study it came from.
• Designed the six-factor evidence scoring model and the public changelog behind it, so no recommendation ever appears without the papers underneath it. 23,000+ grade changes recorded and explained to date.
• Shipped an entity information model connecting every intervention to its mechanisms, biomarkers, and outcomes: 118 biomarkers with region-specific reference ranges, 77 mechanisms, 32 graded outcomes.
• Built the personalisation layer: blood panel ingestion that reads lab PDFs with a vision model and corrects results for draw time against the user's wake anchor, plus Oura, WHOOP, and Withings integration for daily readiness context.
• Operate eleven specialised review agents over the corpus and codebase, covering paper curation, retrieval quality, health-claim compliance across EU and US regimes, and security.
• Shipped the Evidence Assistant, a RAG assistant that answers from the claim database and cites the underlying papers, plus a B2B practitioner tier, an Expo React Native app, and localisation across 3 languages and 7 markets.

Stack: Next.js 16, TypeScript, Supabase, pgvector, Anthropic Claude, Vercel, DeepInfra.

• Designed and developed a multi-tenant SaaS platform enabling organizations to build their own knowledge bases and chat with brand-customized AI assistants (white-label approach with dynamic branding per organization).
• Implemented a scalable RAG architecture with a GPT-4o tool-use loop, hybrid semantic search, and strict tenant isolation at database and search index level.
• Built persistent, project-like chat sessions including a streaming API (SSE), multilingual support, and speech input/output (STT/TTS).
• Delivered the cloud infrastructure as Infrastructure-as-Code, fully automated per-customer CI/CD pipelines, and an onboarding process for new tenants.

Technologies used: Python, FastAPI, Pydantic (v2 noted), Next.js, React, TypeScript, Tailwind CSS, OpenAI / LLMs (GPT-4o), Azure AI Search, Cosmos DB, Azure Blob Storage, Azure Cognitive Services Speech, Azure App Service, Azure Container Registry, Retrieval-Augmented Generation (RAG), Server-Sent Events (SSE), Docker, Terraform, GitHub Actions, REST, OpenID Connect (OIDC), Multi-Tenancy

Built Python-based AI microservices and integrations for an AEC/VR Unity-based SaaS app, focusing on LLM/VLM capabilities, retrieval-backed systems, RESTful APIs, containerized deployment, and an automation microservice for the CAD-to-Unity pipeline.

• Developed a custom Hybrid A* based algorithm in C# to simulate hospital scenarios and detect early-stage design conflicts from collision/spatial data and generate structured reports.
• Solved and automated the time-consuming problem of converting CAD files to usable Unity environments with a custom-engineered and real-time pipeline using a ZeroMQ-based communication layer to distribute workloads across multiple processes and achieve real-time performance.
• Built a Dockerized FastAPI pipeline for CAD-to-Unity automation, combining vision-based object matching, image embeddings, and precomputed metadata to automatically map CAD objects to Unity behavior scripts, assign properties, and reduce repeated AI inference calls.
• Created documentation and examples to help technical users understand, configure, and extend the AI automation pipeline.

• Work as a German/English Language Expert evaluating and rating AI model responses for accuracy, reasoning quality, and natural language use at native/C-level proficiency in both languages.
• Perform structured data annotation and transcription tasks, applying detailed guideline-based scoring and edge-case judgment.
• Conduct Visual Quality Evaluation, assessing AI-generated and model-processed images and video for visual artifacts, factual/compositional accuracy, and adherence to detailed guideline criteria.
• Evaluate and annotate Text-to-Speech (TTS) model output, assessing pronunciation accuracy, prosody, naturalness, and audio quality against structured guideline criteria.
• Evaluate Speech-to-Speech (STS) model interactions, rating conversational audio for naturalness, tone, latency, and response appropriateness in real-time voice-to-voice exchanges.
• Manage concurrent workloads across several platforms simultaneously, prioritizing by task quality and throughput to meet weekly output targets.

• Expert-supervised AI systems for scientific reasoning, model evaluation, and research workflows.
• Built the IAS Problem Solver, an orchestrated system for difficult mathematical reasoning; it achieved 84% in one submitted answer set on the Leipzig mathematics benchmark.
• Built a resumable state-machine pipeline for research-grade mathematics benchmark generation: source selection, LLM-agent-based phenomenon discovery, task synthesis, gold-answer and certificate generation and validation, probing, repair, human feedback, and quality gates, targeting tasks that are difficult, natural, verifiable, and cost-effective.
• Current work extends this into budget-aware AI research workflows for real scientific problems with expert review.

Tech stack: Python, OpenAI/OpenRouter-compatible APIs, embeddings, RAG, SQLite.

• Responsible for the concept, planning, and implementation of the product development of GetMyInvoices 2.0 and the subcomponent InvoiceRails
• Independent work on all aspects of the project, including concept, specification in tickets, and coordination of developers
• Creation, management, and prioritization of tickets to ensure all tasks are completed on time and with high quality
• Carrying out and/or coordinating tests and ensuring the proper implementation of the developed features and functionalities
• Close collaboration with developers to clarify technical requirements and ensure the implementations match the specifications
• Regular reporting on project progress and documentation of key decisions, changes, and risks
• Taking on the subject matter lead for all topics around e-invoicing and Peppol, especially in relation to the InvoiceRails component
• Internal consulting and knowledge sharing on e-invoicing and Peppol for other teams and departments
• Tracking market trends and new developments in e-invoicing and Peppol to continuously adapt the product strategy
• Ensuring the long-term scalability and flexibility of the products for future technical and regulatory changes in the e-invoicing area

• ▶ Built real-time voice assistant (STT → LLM → TTS pipeline) benchmarking and evaluating multiple STT providers including faster-whisper and Azure Speech. achieved sub-3s latency, Groq API (Llama 3) with multi-turn memory - directly handling edge cases in dictation, names and passcode recognition.

• Synthesised 15+ SME case studies on AI-adoption barriers into a structured strategic analysis, and co-organised three startup events within Europe's largest regional high-tech network (1,400+ member companies).

• Responsible for concept, camera, editing, animation, and grading for corporate and B2B productions
• Managing projects from pricing through shooting and post-production to delivery
• Since 2023, a continuous AI-supported production pipeline: Runway, Kling, Veo, Sora, and Seedance for image and moving image content
• ComfyUI for character consistency, ElevenLabs for voice, HeyGen for avatars
• Building reproducible workflows for scalable social media formats
• Building local LLM infrastructure on my own GPU hardware: Ollama, multi-agent systems, RAG, speech-to-text, and text-to-speech
• Process automation for lead generation, email and API workflows, reporting, and document creation

Conventional Paradigm Test (CPT) – AI Evaluation & LLM Research

Development of an experimental evaluation approach to examine “paradigmatic closure” in Large Language Models — that is, the question of how far LLMs can recognize the basic assumptions, values, and limits of the paradigms within which they generate answers.

Design and testing of an additional approach to classic AI benchmarks that does not primarily measure factual correctness or task performance, but instead examines a model’s ability to recognize alternative perspectives, make implicit assumptions visible, and reflect on the limits of its own answer or interpretation framework.

Focus areas: development of evaluation criteria and test questions · LLM evaluation and comparative model analysis · prompt and response analysis · qualitative classification of model answers · study of epistemic compression and value leakage · benchmark and literature research · development of structured assessment and analysis methods

As part of CPT, existing AI evaluation approaches and benchmarks were analyzed, and a minimalist test protocol was developed that classifies model answers by response patterns such as DIRECT, CLARIFY, PLURALIST, REFUSE, and META-AWARE. TruthfulQA was used as the basis for experimental application and comparison with existing reference answers.

Technologies & Methods: Large Language Models (LLMs) · Generative AI · Prompt Engineering · AI Evaluation · TruthfulQA · Benchmark Analysis · Human-in-the-Loop Evaluation · Qualitative Content Analysis · Research & Literature Review

• Managing projects on site
• Coordinating various stakeholders (Operations, IT, Maintenance, suppliers, service providers)
• Leading technicians and external contractors
• Planning resources, schedules, and budgets
• Quality, risk, and safety management
• Conducting daily status meetings
• Escalation management and problem-solving
• Maintaining project tracking, ticket systems, and KPI dashboards
• Materials and spare parts management
• Preparing reports and project status updates

Environment: Infrastructure, Logistics

• Development of an AI-supported personality analysis based on the institute's SDWA4: online survey, AI-supported and automated evaluation, and email delivery
• Requirements analysis, prototype development, evaluation, derivation of a simplified SDWA4-light analysis, continuous improvement, design, Make automation, deployment

End-to-end responsibility for all products - concept, architecture, development, and production operation as the sole developer; in addition, customer meetings, proposals, and marketing.

Underwriting Copilot - AI assistant for industrial insurance (in production at customer sites)

• Supports underwriters in analyzing industrial insurance submissions - in production use at an industrial insurer.
• Framework-independent RAG architecture with Hybrid Search (BM25 + pgvector) across large, mixed document sets.
• Two-stage evaluation and observability pipeline (code assertions + LLM-as-Judge) that makes answer quality, retrieval accuracy, and citation integrity measurable in a regression-safe way.

Kaiki Menu Analyzer - Data intelligence platform (in production at customer sites)

• Automatically captures and analyzes menu data from around 25,000 German restaurants.
• Scalable 7-container architecture (FastAPI, partitioned PostgreSQL, Redis/RQ) with LLM-supported extraction of structured data from PDF, HTML, and images.
• Full CI/CD pipelines (GitHub Actions), production cloud deployment, interactive dashboards (Dash).

Kaiki GEO Atlas - GEO platform (in production at customer sites)

• Measures brand visibility across five AI engines (ChatGPT, Gemini, Perplexity, Grok, Claude), each augmented with web search, orchestrated as a DAG workflow pipeline (Dispatcher → Sub-workflows → Scoring → Report) with fail isolation.
• 6-container deployment (FastAPI, Celery, Redis, PostgreSQL); LLM cost estimation, PDF audit report, rule-based cross-signal insights (no extra LLM cost).

Data Pipeline & Analytics Platform - competitive analysis in the automotive aftermarket

• Automated data pipeline with gap analysis algorithms and role-based access control; 230+ tests.
• Backend with FastAPI, PostgreSQL, SQLAlchemy.

Product development (actively in progress)

BankingGPT - AI assistant for complaint management in cooperative banking

• Security architecture at the core: no AI draft reaches the customer without human approval - the approval decision is in auditable code, not in the language model (monotonic: the model may escalate, never downgrade).
• Real agentic building blocks, each with its own boundary: the model chooses tools itself through an MCP server (read-only, allowlist, capped, fail-safe); sensitive cases are handed off via an open A2A protocol (JSON-RPC, Agent Card, message/send/tasks/get; client implemented by me) to a separate specialist agent (securities/law), which never lowers the review requirement (pinned by test).
• Evaluation-driven over ten analysis rounds; uncovered a security flaw through independent review and blind tests that nine automated runs had missed.
• Voice AI frontend, responding live: covered cases are answered in the conversation, sensitive ones escalate before generation; response latency < 7 s measured (local GPU STT/TTS).

Stack & production readiness: Python, pydantic-ai, FastAPI/Celery, PostgreSQL/pgvector, FastMCP, fasta2a, Docker; multi-tenant capable (physical vector isolation per tenant), PII encrypted, OWASP-LLM reviewed, 275 tests, CI/CD; vendor-portable (Ollama / EU Cloud Vertex).

After-Sales Assistant - agentic RAG/GraphRAG assistant on public OEM manuals (automotive after-sales)

• Genuinely agentic on LangGraph: ReAct agent with four tools and conversation memory - the model decides on its own whether to use the manual (RAG, Chroma), a knowledge graph (GraphRAG, Neo4j/Cypher - decodes warning lights), or a workshop/booking service.
• Human-in-the-Loop before the irreversible action: before every appointment booking, the graph pauses (interrupt) and gets the driver's explicit confirmation - the same approval-before-action discipline as in BankingGPT, in a different framework.
• Eval as CI gate: a three-part scorecard (RAGAS grounding + deterministic tool-routing accuracy + DeepEval safety: does the answer mention the warning first when there is a critical warning?) blocks the pipeline; provider-agnostic (OpenAI/Azure/Anthropic), FastAPI with token streaming.

Stack: Python, LangChain/LangGraph, Chroma, Neo4j, RAGAS/DeepEval, FastAPI, Docker.