michalwojdylak.com

About

I'm Michał Wojdylak, AI Lead at WPP Enterprise Solutions in Poznań. I build and operate the systems that take machine learning from notebooks to reliable, scalable production services - data, training, deployment and monitoring.

I focus on the parts of AI that have to work at 3am: serving large language models with predictable latency, designing inference platforms that scale with demand, optimizing GPU utilization and cost, and building the tooling that lets teams ship models safely and often.

On this site I publish reproducible benchmarks: I pick a question teams actually face, hold everything else constant, and release the code and raw results alongside the write-up.

Experience

  1. AI Lead · WPP Enterprise Solutions

    Nov 2023 - present

    • Building high-performance AI inference platforms with GPU acceleration, dynamic model loading and distributed serving.
    • Designing enterprise AI solutions, from low-latency real-time inference to cost-optimized serverless architectures.
    • Aligning AI initiatives with DevOps, Product, Backend and Frontend teams.

  2. Senior AI Engineer · WPP Enterprise Solutions

    May 2023 - Oct 2023

    • Technical lead of the AI team building computer vision and LLM solutions.
    • Built and managed AI infrastructure on AWS.

  3. AI Engineer · WPP Enterprise Solutions

    May 2022 - Apr 2023

    • Built a microservice application with Celery, RabbitMQ and Redis.
    • Deployed state-of-the-art and custom models to AWS and Azure.

  4. Junior AI Engineer · WPP Enterprise Solutions

    Nov 2021 - Apr 2022

    • Generated synthetic data and trained and optimized computer vision models.
    • Designed and built end-to-end cloud ML solutions.

Skills

Inference & model serving
GPU-accelerated inference platforms, NVIDIA Triton Inference Server, TensorRT, ONNX Runtime, dynamic model loading, distributed serving1 article on Model Serving
LLMs
vLLM, SGLang, TensorRT-LLM, quantization (FP8, AWQ, GPTQ), Mixture-of-Experts models2 articles on LLM Inference
Retrieval & RAG
Hybrid search on OpenSearch, rerankers, agentic and graph-based RAG, evaluation with LLM judges1 article on RAG & Retrieval
Cloud
AWS, Azure, GCP; low-latency real-time and cost-optimized serverless AI architectures1 article on Cost Optimization
Computer vision
Synthetic data generation, model training and optimization, real-time inference
Backend & leadership
Python, Docker, microservices with Celery, RabbitMQ and Redis; leading an AI team and working with DevOps, Product, Backend and Frontend teams

Education

  • MSc Eng., Automatic Control and Robotics

    Poznań University of Technology · Robots and Autonomous Systems

    2020 - 2021

  • BSc Eng., Automatic Control and Robotics

    Poznań University of Technology · Robotics

    2016 - 2020

Certifications

  • Google Cloud Professional Machine Learning Engineer
  • Microsoft Certified: Azure AI Engineer Associate
  • AWS Certified Developer - Associate
  • Microsoft Certified: Azure AI Fundamentals