Human-in-the-Loop Experts in Berlin

Workflow-automation SaaS for operations teams (Berlin, 120 people); full-time freelance engagement reporting to the CEO: an initial 12-month interim mandate, extended twice through the AI build-out; owned product for one squad and coached the other product managers on process.

• Led generative AI (LLM) integration into the core product: from LLM-powered steps to natural-language workflow authoring and step-level automation suggestions, plus AI-managed dynamic workflows, shipped behind eval gates with human-in-the-loop fallbacks: AI-drafted workflows grew to 31% of all new workflows, and median time-to-first-workflow fell from 3 days to 4 hours.
• Packaged the AI capabilities as a usage-based add-on priced on executed automation steps, working with sales and marketing on positioning: ~€800K added ARR in the first year, and adopting accounts churned 1.8 pp less.
• Owned the roadmap end to end: replaced feature-request-driven quarterly planning with an outcome-based rolling roadmap built on quarterly bets and explicit kill criteria, presented monthly to the executive team and quarterly to the board.
• Rebuilt the product-management operating system: weekly customer-discovery cadence incl. workshop facilitation, RFC/decision-doc reviews and a single quarterly metrics narrative; coached four product managers, one promoted to senior during the engagement.
• Closed the engagement as scoped: hired and onboarded the permanent VP Product, handed over the process playbook and roadmap, and exited on schedule in June 2026.

• Own the 12–18 month roadmap and key outcomes for Elli's enterprise customer platform, covering onboarding, pricing, billing, analytics and broader platform modernization; redesigned the Fleet onboarding funnel to double conversion, supporting a projected €20.7M revenue uplift by 2028.
• Lead the broader Energy Intelligence product and directly own its AI/ML, asset and portfolio-optimization capabilities, including MLOps and safe strategy deployment, strategy lifecycle management and backtesting; delegated data and V2G integration roadmap ownership to a new PO as the platform scope expanded.
• Built a Human-in-the-loop GenAI/RAG support workflow, increasing L1 resolution by 24%, routing accuracy to 91%, and reducing L2 workload by 30%.
• Introduced standardized data contracts and a self-service Python toolkit for traders and Data Scientists, increasing platform adoption by 15% and reducing support effort by 50%.
• Built and scaled a real-time orchestration product from 32 to 3,000+ endpoints across four markets, growing recurring revenue from €1.4k to €56.3k MRR.
• Developed product and AI capability across the organization, training 20 PMs on RAG, agents and prototyping; mentoring a junior PM and coaching an Enterprise Platform Tech Lead toward Product Management ownership.

• Root cause analysis of complex business, technical, and data-related errors in PowerCloud across process, booking, and system boundaries.
• Data-driven management of payments and receivables; contributed to reducing historical receivables from over 100 Mio. EUR to under 25 Mio. EUR.
• Identification of automation and straight-through processing potential at the interface between business departments, IT, and external service providers.

Following the Vaillant engagement, established independent practice focused on senior SAP advisory and building ASTRA — a methodology, not a tool. ASTRA is independent and unbiased: the same four pillars serve CIO / Business Leadership, SI / SAP Practice, and Programme Director / Principal with equal weight. No audience is primary.

ASTRA serves three audiences with equal weight — CIO / Business Leadership · SI / SAP Practice · Programme Director / Principal. ASTRA is independent and unbiased — it does not sit inside any SI practice and does not serve any single audience as primary customer. Each audience receives a role-calibrated output from the same analysis.

One methodology, two outcomes:

OUTCOME 1 — SAP TRANSFORMATION GOVERNANCE:

Any active SAP programme: ECC→S/4HANA Greenfield, Brownfield, Global Rollout, AMS. The four pillars answer the four questions that determine whether a transformation succeeds — for any of the three audiences.

• Are business requirements clean — or is the SI building what SAP standard already covers? (Functional Auditor → all three audiences)
• Is the custom object inventory healthy — or accumulating debt that blocks upgrades? (RICEFW Auditor → Director and CIO)
• Is the SteerCo reporting accurate — or is leadership seeing a Watermelon? (Truth Engine → CIO and Director)
• Are people ready for the system they are receiving? (Human Signal → CIO and Director)

OUTCOME 2 — AUTONOMOUS ENTERPRISE READINESS:

Any programme deploying Joule, pursuing RISE, or planning an AE migration. The same four pillars answer the questions SAP does not — again, calibrated per audience.

• Which custom objects block the 224 Joule AI agents? (RICEFW Auditor → SI gets remediation paths, Director gets verdicts, CIO gets financial impact)
• Is the Fiori deployment complete enough to activate Joule? (Truth Engine → all three audiences)
• Is the workforce trained for AI-native working? (Human Signal → CIO and Director)

Four pillars (each serves all three audiences):

• Functional Auditor: Clean Core Score 0–100. Section A (CIO): KG fidelity risk in €. Section B (Director): SI challenge questions, man-day saving. Section C (SI): Language Translation Ledger, SAP alternatives.
• RICEFW Auditor: Joule API Pathway Audit, Clean Core Level A–D, Technical Debt Index 1–10. Section A: financial impact of blocked AI agents. Section B: RETIRE/MIGRATE/REDESIGN/KEEP verdicts. Section C: BTP remediation paths.
• Truth Engine: Watermelon Detection, Fiori Deployment check, Programme Truth Score 0–100. Section A: 3 SteerCo questions. Section B: 48-hour action list. Section C: vendor friction map.
• Human Signal: three-stage Fiori adoption model, HITL Policy readiness, Human Signal Score 0–100. Section A: GUI training weight, AE adoption financial risk. Section B: go-live recommendation, stakeholder resistance map.

On-Demand SAP Advisory:

• SAP Programme Management — end-to-end governance, factory management, multi-stream delivery, vendor oversight, SteerCo reporting for Greenfield, Brownfield, Rollout and AMS programmes.
• Delivery Management — RICEFW oversight, Data Migration governance, Release & Cutover management, Hypercare, SLA/KPI management.
• Due Diligence & Landscape Assessment — SAP landscape review, AS-IS/TO-BE roadmap, RFP structuring, SI selection advisory.
• IT Strategy & CIO Advisory — transformation roadmap, Clean Core strategy, Autonomous Enterprise readiness planning.
• Programme Recovery — distressed programme assessment, rapid governance setup, escalation resolution.

Operating through senior professional networks across EMEA.

• Designed and deployed enterprise Retrieval-Augmented Generation (RAG) applications using LangChain, LangGraph, vector databases, embeddings, and open-source LLMs served through vLLM on GCP GPU infrastructure.
• Built agentic AI workflows using LangGraph with planning, reasoning, tool execution, persistent memory, session management, and Human-in-the-Loop approval mechanisms.
• Developed LLM-powered automation systems integrating BigQuery, SQL pipelines, and external advertising APIs including Meta, TikTok, Amazon, Snapchat, Google, and Pinterest, reducing manual operational workflows.
• Architected multi-agent AI systems for enterprise analytics and decision-support workflows, enabling autonomous task execution and intelligent data interactions.
• Implemented retrieval optimization strategies including multi-retriever architectures, semantic search, context optimization, and query improvement techniques, improving response relevance by approximately 40%.
• Engineered structured prompting strategies, function-calling schemas, and validation workflows to improve reliability of multi-step LLM applications.
• Designed scalable AI services using Python, FastAPI, Cloud Run, Pub/Sub, BigQuery, Docker, and cloud-native deployment architectures.

OpsWorker turns Kubernetes alerts into root-cause analyses, on top of the monitoring a team already runs. I lead the technical side: the agent architecture, the AWS infrastructure it runs on (fully inside EU regions), and the engineering decisions behind it, read-only in the cluster by default, human in the loop for judgment. The stack underneath: Amazon Bedrock and Bedrock AgentCore, agents built with the Strands Agents SDK, the Claude and OpenAI APIs, and the Kubernetes API.

• Lecturer for the GenAI Track at the Master School Institute of Technology
• Development of a full-stack AI application (React + Python/FastAPI) for automated supplier product import with intelligent column and category classification (4-layer hierarchical) including human-in-the-loop validation

• Leading end-to-end UX for two AI SaaS products in closed beta, including LLM-interaction design, prompt-UX, and human-in-the-loop patterns with commercial distribution signed for launch in Q3 2026
• Built a self-built LLM reframing and RAG-correction pipeline powering multi-profile CV and case-study generation in production use
• Shipping real code alongside research, including Three.js/GLSL portfolio work, Figma-API tooling, and a Chrome MV3 extension for session-sync automation

• Designed a compact framework to generate prompt–response sets for CRM lifecycle scenarios (onboarding, activation, retention, reactivation).
• Included adversarial variants (ambiguous requests, conflicting instructions, policy traps).
• Created a scoring rubric for factuality, tone, and coherence.
• Developed a lightweight guideline for annotator alignment and disagreement resolution.

• Own full-stack development of an AI-native enterprise platform built on TypeScript, React, Vite, tRPC, Hono, and PostgreSQL, delivering AI-powered consulting workflows to B2B clients.
• Designed and shipped a multi-agent AI system using ReAct framework and Claude skills-style workflow patterns, including an intelligent PM assistant with rich system prompts, slash commands, tool integrations, and streaming chat UI.
• Architected an LLM evaluation framework: rubric-based LLM-as-judge, golden datasets, regression testing, and automated quality gating — ensuring consistent AI output quality at scale.
• Integrated LangFuse for end-to-end LLM tracing, conversation replays, and evaluation pipelines, enabling data-driven prompt optimisation that reduced token costs and response variance.
• Built with Drizzle ORM, pgvector, and knowledge graphs for structured data access, semantic search, and relationship-aware AI reasoning across the platform.
• Led TanStack React Query migration across the application — replacing manual state management with centralised caching and automatic refetching, reducing data-fetching boilerplate significantly.
• Practiced AI-native development throughout: Claude Code, Codex, Perplexity SDK, and LLM-assisted testing across the full development lifecycle. Deployed on Vercel + Azure ACA with Biome for linting/formatting.

• Shipped Memorable, a production iOS app using on-device CLIP-based semantic photo search. Owned the full stack: Core ML conversion, local inference pipeline, App Store release, and post-launch iteration.
• Built a practical AI deployment framework that covers workflow redesign, use-case prioritization, system integration, eval planning, and human-in-the-loop controls.
• Conducting AI use-case discovery and advisory conversations with professionals in legal, tax, and real estate sectors.

• Designed AI-powered workflows that transform complex tasks into intuitive, scalable solutions
• Combined UX strategy, technical implementation, and systems thinking to simplify processes and empower users
• Conducted user research and usability testing to develop a practical, accessible AI guide for family caregivers
• Performed data and competitive analysis to inform content and feature development
• Created wireframes, mockups, interactive prototypes, MVP, design systems, and styleguides
• Executed UX and accessibility audits to reduce manual effort and streamline processes
• Scaled content production across multiple formats through UX writing and automation
• Managed projects and design operations, mentoring team members
• Technologies and methods: ChatGPT, Claude, OpenAI API, Make, n8n, ElevenLabs, HeyGen, TTS, HITL design, prompt engineering, Figma, Flutterflow, Miro, Notion, design thinking, workflow mapping, business process mapping

• Defined and executed the end-to-end vision for Allride’s core mobility product, launching a fully functional mobile app on iOS and Android in under 3 months.
• Led product from 0 to 1 crafting roadmap, aligning teams, and driving execution to meet user needs and business objectives.
• Built a tiered S/M/L/XL subscription model based on mobility usage, launching the first MVP with built-in rewards and coupons to drive adoption of Allride’s recurring plans.
• Designed and implemented a high-conversion referral program that rewarded both referrers and invitees, driving 12% of new user acquisition through organic growth loops.
• Developed and launched Allride for Work, a B2B mobility benefit solution that enabled companies to offer sustainable commute plans to employees, driving corporate adoption and unlocking a new recurring revenue stream.
• Aligned product initiatives with sustainable mobility goals, contributing to rapid growth and early media coverage.

• Leading the product lifecycle for sovereign AI platform and operating system teams for enterprise & government clients and internal stakeholders (infra, solution delivery, support, revenue)
• Built and scaled the platform from a 200-user beta to a full rollout of 70K+ members at the Bundesagentur für Arbeit (BA), secured with ISO 42001 and EU AI Act compliance
• Architected the shift to a multi-tenant shared inference, increasing GPU cluster utilization from 20% to 85% and reducing infrastructure cost-to-serve by 40% for SaaS clients
• Shipped model quantization, allowing clients to run advanced LLMs on legacy hardware (A100s GPUs) instead of the H100s, saving upwards of 70% cost per enquiry
• Abstracted complex Helm configurations into a dynamic model manager, reducing the time to install or swap models by ~80%
• Killed an expensive move to build own dashboard service, pivoting to an API-first data strategy that clients can consume directly and saving €100Ks in opex and capex
• Built a safety-first agent marketplace and control plane lighthouse project for a Tier-1 bank, allowing internal teams to deploy autonomous agents within strict regulatory guardrails