AI Architect& Engineer

Berlin, Germany

I build enterprise AI systems and co-lead a portfolio of reusable AI solutions.

My work combines hands-on engineering, architecture decisions and customer collaboration, with a focus on agentic AI, RAG and evaluation.

Contact

Selected work & responsibilities

Associate Manager · Accenture · Financial services

Enterprise AI architecture & delivery

Technical leadership · Hands-on engineering

Retrieval and agentic systems are central to my enterprise AI work in financial services.

My contribution

I lead technical scoping and architecture discussions and translate business, integration, security and regulatory requirements into application designs. I stay hands-on throughout implementation, production rollout and ongoing operation.

Regulated delivery

My work has included PII and sensitive-data handling, access and usage tracking, AI risk requirements, data residency, zero-retention patterns and private or on-premises deployment constraints.

Center for Advanced AI · Accenture

AI asset portfolio co-lead

Portfolio co-leadership · Steering committee contribution

Alongside client delivery, I co-lead the AI asset portfolio at Accenture's Center for Advanced AI. The portfolio brings together reusable AI solutions, documentation and demonstrations for teams to build on across client engagements.

My contribution

I develop and curate assets so teams can find, understand and reuse them, with attention to maturity and documentation. Working with delivery and account teams, I identify suitable capabilities for client opportunities and help colleagues apply them.

The portfolio lives inside a digital brain and uses Open Knowledge Format (OKF) to structure knowledge about available solutions. This helps teams find, understand and reuse the right assets.

RAG, agentic systems & evaluation

My focus is on building enterprise AI applications around knowledge retrieval and agentic workflows, evaluating their technical quality and improving their performance and adoption in production. Much of this work happens in regulated environments, where privacy, security, data residency and governance requirements directly shape architecture and deployment decisions.

Knowledge & retrieval architectures

My retrieval work brings together multiple knowledge sources and document processing, with implementation choices guided by the information users need and the constraints of each use case.

AI applications & agentic systems

I design and build end-to-end AI applications, from backend APIs and agent orchestration to user-facing interfaces. Using Python, FastAPI and LangGraph, I connect retrieval, tools and MCP integrations to real business workflows, while designing around constraints such as sensitive-data handling, data residency, retention requirements and private deployment models.

AI evaluation & reliability

I evaluate AI systems on real business tasks and build self-service evaluation tools that enable client teams to run their own assessments. Using evaluation frameworks, I measure answer quality, task completion, reliability, latency and cost to help teams uncover failure modes, catch regressions and make informed architecture and deployment decisions.

Validation, rollout & adoption

I work with business teams to validate AI applications, roll them out into production and improve them in ongoing use. I combine user feedback with performance KPIs, adoption rates and operational stability metrics to assess how applications work in practice and prioritise improvements.

Contact

hello@lucapankratz.ai Connect on LinkedIn