de

[ hello & welcome ]

I'm Jan Carstens,AI engineer., LLM engineer., agentic-AI engineer., solution architect., your sparring partner.

I identify the challenges in your business and use the right technology to solve them. As an industrial engineer, I connect business and technology where it counts: measured by impact and ROI, from solution concept to systems running in production.

Portrait of Jan Carstens

[ Industry experience ]

  • Energy & Utilities
  • Insurance
  • Manufacturing
  • Machinery & Plant Engineering
  • Automotive & Mobility
  • Healthcare
  • Chemicals
  • Banking & Financial Services

[ profile ]

Your challenge meets solution expertise.

In most companies, hidden potential lies dormant in many places. What holds implementation back is rarely the idea; it is the solution space: which technology fits, and how it reaches everyday operations. That's exactly where I come in: as a management consultant, I develop the solution concepts and put the right technology in the right place. And as a solution engineer, I help build the systems that turn it into results.

What I bring: as an industrial engineer (M.Sc.), I have spent my entire career resolving process bottlenecks through optimization and automation. Working with me gets you both: the concept that scales, and the hands that deliver.

[ services ]

Four focus areas, one standard: measurable impact.

[ consulting ]

Analyzing Your Potential

Identifying, evaluating, and prioritizing your potential. Because not every challenge should be tackled with AI, only where technology creates measurable value in your day-to-day business.

  • Analysis of the challenges in your day-to-day business
  • Deriving potential that the use of technology can realize
  • Prioritizing potential by effort and impact
  • Complete implementation roadmap including processes, governance, and roles

[ concept ]

Developing Solution Concepts

Solution concepts that fit and grow: tailored to your use case, embedded into your existing system landscape, and designed for scale and a fast ROI from day one.

  • Vendor-neutral solution architecture
  • Accounting for requirements and restrictions of your existing infrastructure
  • Designed for scale and fast ROI
  • Fast path to demonstrator, proof of concept, or MVP

[ development ]

Realizing the Potential

Together with the colleagues who face the challenge every day, I build the solution iteratively: I create the tools your processes need and bring them into productive use.

  • Make: developing custom tools on proven technologies
  • Buy: selecting and introducing established solutions
  • Connecting proprietary systems and legacy software
  • Extending existing integrations
  • Testing, deployment, and documentation

[ scaling ]

Scaling & making it stick

Away from island solutions, toward capabilities across the entire company: what was built for one use case becomes a building block for the next and grows into a shared technological base.

  • Reusable components: system integrations and infrastructure carry the next use case
  • Lower costs and shorter delivery time with every additional use case
  • Enabling your people so solutions live on in daily work
  • A clear operating model so solutions are carried after the project ends

[ projects ]

Selected references.

[ erp × agents ]

Agentified ERP system

delivered

An ERP system fully exposed as an MCP server, operated by AI agents instead of humans. Employees talk to a supervisor agent that applies the right skills per task and spawns subagents on its own.

impact

Administrative work by natural language instead of click paths. Operations get faster because end-to-end processes run automated without manual steps in between.

Read the full reference

[ forecasting × artisan food ]

Demand forecasting for artisan food producers

delivered

Forecasting platform that unifies sales data from stores, webshop, and vending machines into one source of truth and predicts daily demand per article, with a reliable uncertainty band.

impact

Production and procurement planned on numbers instead of gut feeling: less overproduction and write-offs, better availability.

Read the full reference

[ geodata × site selection ]

AI-assisted site analysis

delivered

Data-driven site search for unstaffed sales containers: the search area is rastered without gaps and scored on six weighted criteria from open geodata. The result is not a report but an interactive system that drives the selection live in the workshop.

impact

Site decisions based on data instead of gut feeling. Only AI-assisted development made this depth of analysis affordable, and the system carries further sites and other businesses.

Read the full reference

[ transcription × sales ]

GDPR-compliant conversation transcription

delivered

Live transcription for sales: conversations are transcribed in the browser without any audio being stored, and an AI agent turns the transcript into a visit report, tasks, and CRM entries.

impact

Sales holds conversations instead of writing minutes: minutes after the meeting, report and tasks are in the CRM, GDPR-compliant without a recording.

Read the full reference

[ agent platform × governance ]

Agent platform

delivered

No-code platform where business users build and visually orchestrate their own AI agents, including a governance layer, approval guards, and a human-in-the-loop inbox.

impact

AI scales across departments without losing control: every agent action stays traceable, and critical steps require human approval.

Read the full reference

[ deep research × sales ]

Go-to-market deep research

delivered

Deep-research agents analyze existing customers and derive prioritized AI use cases (farming), or turn a search profile into a validated target list with a cited verdict per company (hunting).

impact

Sales walks into every conversation prepared: individually researched talking points instead of generic pitches, in active use every day.

Read the full reference

[ knowledge graph × insurance ]

Knowledge graph for insurance advice

delivered

A knowledge platform that understands insurance policies: a knowledge graph links life situation and product world, and an advisory website that adapts to the visitor is generated automatically, with cited answers.

impact

A folder of documents becomes an advisory journey: structured content, answers with source citations, and prequalified inquiries for the broker to close.

Read the full reference

[ approach ]

Technology that actually lands in the organization.

Full integration

AI not as an island solution but as a capability in your company: fully integrated into your processes, with the data and systems you already have, used by your people in their daily work.

Governance from day one

Roles, processes, responsibilities, skills, and audit keep technology anchored and secured instead of fading out after the pilot.

No-code, low-code & code

Flexibility in how your solution is built: from no-code to custom development, from make to buy. That realizes the full potential without the limits of off-the-shelf tools, and without reinventing the wheel.

Value over hype

As an industrial engineer, I put business value first: without concrete value for the business, AI remains technology without impact, and hype without substance.

[ technology ]

The tools I build with.

Agentic Coding

Coding agents support me in specifying, verifying, and documenting code. The result: fast demonstrators and clean, well-documented tests.

  • Visual Studio Code
  • Claude Code
  • Cursor

Agents & Language Models

I use the full spectrum from no-code to low-code to code, so realizing your potential never fails at a tooling limit.

  • LangGraph
  • AG2 / AutoGen
  • CrewAI
  • OpenAI Agents SDK
  • Anthropic Agents SDK
  • Hugging Face
  • Ollama
  • vLLM
  • LM Studio
  • Perplexity
  • MCP
  • n8n

Machine Learning & Natural Language Processing

Not every problem needs a language model. Where it fits, I rely on classical machine learning, deep learning, and natural language processing.

  • TensorFlow / Keras
  • scikit-learn
  • spaCy
  • NLTK
  • BERTopic

Data Engineering & Data Architectures

The basis of every reliable AI application is its data foundation: from tabular to vector to document and graph databases, plus the caching and queueing that carry background processing.

  • PostgreSQL / pgvector
  • Elasticsearch
  • Neo4j
  • Redis
  • Celery

Cloud & Operations

Scalable, low-maintenance operations require clean architecture, logging, monitoring, and CI/CD from the start.

  • Azure
  • Docker / Kubernetes
  • GitLab
  • Azure DevOps
  • LangFuse
  • Prometheus

Frameworks

Frontend and backend frameworks that speed up development, without cutting corners on security and scalability.

  • Python
  • TypeScript
  • FastAPI
  • Django
  • Next.js / React
  • Vite

Requirements I Meet

Secure operation takes more than technology: I account for these requirements from concept to production.

  • GDPR
  • EU AI Act
  • ISO 27001

[ contact ]

Let's exchange experiences.

Your company, too, has untapped potential lying dormant. Often all that's missing are the right tools to realize it. Whether it's a first idea, a concrete use case, or simple curiosity: reach out, no strings attached. I look forward to the exchange.