Principal AI & MLOps Architect
Piotr Łączkowski
From Research Labs to Boardrooms — End to End
- Private & Public · Scale-Up & Enterprise · PhD
- Architectural Clarity Meets Hands-on Engineering
- CNRS · Thales · BackMarket · Vestiaire · ADEO
I like figuring things out — taking complex, ambiguous problems and turning them into reliable production systems. From public research labs (CNRS, CEA, Thales) to European scale-ups (BackMarket, Vestiaire Collective) and global enterprise (ADEO, 150K+ employees), I combine strategic architectural advisory with hands-on engineering across 200+ production AI systems and open-source frameworks. Thinking and making belong together.

- Business Value
- €500M+
- Systems Built
- 200+
- Open Source Libraries
- 7+
- Years Building AI
- 17+
Cumulative across research labs, scale-ups and enterprise roles. Only selected use cases are shown; most enterprise work is under NDA.
About Me
Thinking and Making Belong Together
Good work starts with understanding — not every problem needs AI, and not every new technology is useful. From public research labs to executive boards, I combine architectural clarity with hands-on implementation across scale-ups, unicorns, and global enterprises.
- PhD in Physics
- 7+ Open Source Libraries
- Principal AI & MLOps Architect
- Enterprise & Scale-Up Track Record
Thinking & Making
Advisory and hands-on engineering belong together — 200+ production AI systems and 7+ open-source frameworks built end-to-end.
Pragmatic AI & GenAI
Good work starts with understanding. From RAG/GraphRAG and agentic workflows to classical ML — choosing what actually fits the problem.
End-to-End Architecture
Python, Go, React, TypeScript, Kubernetes — architecting complete systems across data pipelines, backends, and infrastructure.
Staying Current
Continuous study and hands-on testing of evolving models, tools, and engineering practices — separating signal from noise.
Staying Close to Technological Change
Technical Scope & Architecture Domains
Continuous hands-on exploration and production delivery across 150+ technologies — evaluating what holds up in practice across the complete AI/ML lifecycle.
GenAI & LLMs
LLMs, RAG and agentic systems
- OpenAI / GPT-4
- Claude / Anthropic
- Gemini
- Llama 3
- Mistral
- LangChain
- LangGraph
- LlamaIndex
Machine learning
Classical and deep learning
- PyTorch
- TensorFlow
- Keras
- JAX
- Scikit-learn
- Transformers
- BERT
- GPT
MLOps & LLMOps
Production ML at scale
- MLflow
- Kubeflow
- Weights & Biases
- DVC
- BentoML
- Model Registry
- Feature Stores
- Feast
Backend & APIs
Scalable systems
- Python
- FastAPI
- Django
- Flask
- Pydantic
- GoLang
- Gin
- gRPC
Cloud & DevOps
Infrastructure and operations
- GCP
- AWS
- Azure
- Vertex AI
- SageMaker
- Kubernetes
- Docker
- Helm
Frontend & tools
Interfaces and productivity
- TypeScript
- React
- Next.js
- Node.js
- Streamlit
- Gradio
- Material UI
- TailwindCSS
Portfolio
Selected work
Enterprise AI systems, agentic architectures, and reusable open-source frameworks built from research to production.
PlatformsLicensing & Observability HubWatchtower Platform
Centralized licensing, entitlements, and runtime observability platform engineered with Go, React, Redis, and Kubernetes.
- GoLang
- React
- TypeScript
- PostgreSQL
- +2
PlatformsAI Agent OrchestrationAIgency Platform
End-to-end AI agent and workflow orchestration system supporting multi-step reasoning, RAG, and human-in-the-loop governance.
- GoLang
- React
- TypeScript
- MongoDB
- +3
PlatformsAI Knowledge & Blueprint SystemGalaxy Architecture Hub
Centralized engineering repository and architecture hub for structuring, evaluating, and sharing reusable AI/ML system blueprints.
- GoLang
- React
- TypeScript
- PostgreSQL
- +2
AI/MLRAG & Graph-RAG(opens in a new tab)Chatbots - LLM Based Systems
Production conversational and knowledge-retrieval systems engineered with hybrid RAG, Graph-RAG, semantic reranking, and Langfuse observability.
- LLM
- LangChain
- LlamaIndex
- Ollama
- +3
Open SourceOpen Source Library(opens in a new tab)Keras Data Processor (KDP)
Advanced data preprocessing library based on Keras preprocessing layers. Standalone or as first layers.
- TensorFlow
- Python
- GitHub
- Poetry
AI/MLE-commerce AI System(opens in a new tab)Phoenix Recommendation Engine
Unified recommendation model combining Collaborative Filtering, Two-Tower Retrieval, and Ranking with dynamic weighting, multi-task learning, and feedback loops.
- Keras
- TensorFlow
- Transformers
- Python
- +1
Capabilities
Consulting & Engineering Services
I partner directly with engineering and leadership teams to design, build, and scale reliable AI systems — connecting strategic clarity with hands-on implementation from architecture to production.
Consulting & engineering
AI Strategy & Architectural Advisory
Clarify where AI creates genuine value, evaluate technical feasibility, and design pragmatic roadmaps grounded in engineering reality.
Key Deliverables
- Architecture review
- Strategic roadmap
- Feasibility & ROI analysis
- Technology selection
GenAI & Agentic Systems
Design and build production-grade LLM applications, RAG/GraphRAG pipelines, and multi-agent workflows with rigorous evaluation.
Key Deliverables
- Custom LLM systems
- RAG / GraphRAG
- Agentic workflows
- Evaluation & guardrails
MLOps & AI Infrastructure
Bridge the gap from prototype to dependable production with scalable pipelines, observability, and enterprise-grade reliability.
Key Deliverables
- ML pipeline architecture
- Production deployment
- Monitoring & drift detection
- Cost & latency control
Team Enablement & Coaching
Strengthen your organization's internal engineering capabilities through hands-on workshops, architectural coaching, and mentorship.
Key Deliverables
- Capability assessment
- Engineering workshops
- Architecture coaching
- Knowledge transfer
How We Work
Different Ways to Work Together
Whether you need a complete system delivered end-to-end, an embedded technical partner, or a focused architectural review, the collaboration format adapts to your team and context.
Project-Based Delivery
You have a defined problem or opportunity. I scope, architect, and build the solution from initial design to production deployment and handover.
Embedded Technical Partnership
You have an ongoing stream of technical challenges. I integrate directly with your engineering and leadership teams on a sustained basis.
Focused Expert Input
You need a sharp architectural audit, feasibility assessment, or technology review. Short, high-density engagements that deliver clear answers.
Training & Enablement
You want your team to build stronger capabilities in AI, GenAI, or MLOps. Hands-on workshops and coaching tailored to your actual stack.
Products & technology
Alongside client work, I build and maintain open-source libraries under UNICOLAB — for data preprocessing, Keras models, ML pipelines and agent workflows.
See the open-source libraries- KerasFactory
- MLPotion
- flowyml
- KDP
- GoLangGraph
Have a Complex Technical Problem Worth Figuring Out?
Let's start with a conversation. Work directly with me on architectural strategy, end-to-end production engineering, or technical team enablement.
Career Journey
17+ Years of Engineering & Research
From experimental physics research to leading data and AI teams, and now independent AI engineering and advisory. Most recent first.
Work
- 2025 – PresentNOW
Principal AI & MLOps Architect
Independent Consultant · Paris, France
Hands-on AI architecture, production GenAI & agentic systems engineering, and MLOps infrastructure delivery for enterprise clients, backed by 7+ open-source frameworks.
- Partnering directly with enterprise and scale-up teams on AI strategy, architecture review, and end-to-end production delivery
- Architecting and deploying production RAG/GraphRAG systems, autonomous agent workflows, and scalable MLOps platforms
- Author and maintainer of 7+ open-source frameworks (GoLangGraph, KDP, FlowyML, MLPotion, KerasFactory) that accelerate client engineering
- Full-stack system implementation across Python, Go, React/TypeScript, and Kubernetes
- GoLang
- React/TypeScript
- LangGraph
- RAG
- +3
- 2022 – 2025
Head of AI
Leroy Merlin · Paris, France
Architected and shipped 15+ AI/ML systems: demand forecasting, fraud detection, GenAI chatbots, recommendation engines.
- Built end-to-end MLOps platform on GCP/Vertex AI — model registry, A/B testing, monitoring
- Personally coded GenAI systems: RAG chatbots, marketing image generation (Stable Diffusion), negotiation coach (LLM)
- Architected demand forecasting & supply chain optimization delivering €60M+ yearly savings
- Python
- GCP
- Vertex AI
- LangChain
- +3
- 2019 – 2022
Lead Data Scientist
Vestiaire Collective · Paris, France
Built computer vision systems, NLP pipelines, and pricing engines from prototype to production on AWS.
- Built reverse image search engine (TensorFlow, ElasticSearch) — 89% match accuracy, <1s latency
- Coded pricing engine with Bayesian regression on SageMaker — 45% margin improvement
- Designed and deployed MLOps infrastructure: CI/CD for ML, model monitoring, feature stores
- TensorFlow
- AWS
- SageMaker
- Python
- +3
- 2016 – 2019
Head of Data/Full Stack Developer
BackMarket · Paris, France
Built fraud detection, product ranking algorithms, and data infrastructure. Full stack: Python/Django + data pipelines.
- Coded real-time fraud detection preventing €2M+ losses — TensorFlow autoencoders
- Built product ranking algorithm increasing CTR by 35%
- Architected data platform: Airflow pipelines, Snowflake warehouse, K8s deployment
- Python
- Django
- TensorFlow
- Airflow
- +3
- 2012 – 2016
R&D Engineer/Researcher
CNRS/Thales · Palaiseau, France
Nano-fabrication of Spintronics transistors, data analysis automation and modeling software.
- Authored data automation software and analysis pipelines used across research labs
- Python
- Data Analysis
- Statistical Modeling
- Automation
- +1
- 2012 – 2016
R&D Engineer
CEA · Grenoble, France
Built data analysis and visualization tools for nano-fabrication research.
- Python
- Data Analysis
- Visualization
- Physics
- 2006 – 2008
R&D Engineer
CNRS · Grenoble, France
Engineering automation software for near-field microscopy.
- Automation
- Microscopy
- Physics
Education
- 2009 – 2012
PhD in Physics
Joseph Fourier University · Grenoble, France
Spin currents and spin Hall effect in metallic lateral nano-structures.
- Published 5 papers in peer-reviewed journals
- International conference presentations
- Spintronics
- Data Analysis
- Quantum Physics
- Research
- 2003 – 2009
Master in Physics
Adam Mickiewicz University · Poznan, Poland
Experimental Physics and NanoTechnology for Spintronics.
- Graduated with honors
- Research in spintronics & nanotechnology
- Experimental Physics
- Nanotechnology
- Spintronics
- Data Analysis
Piotr Łączkowski