Max Rehman Linder

Ericsson Stockholm

Fullstack AI Researcher Sep 2025 — present

Building dataset, training, and evaluation pipelines for a proprietary large language model, from continued pre-training through reinforcement learning.

Thesis Worker, System Comprehension Lab Jan — Sep 2025

Master’s thesis on vocabulary expansion of LLMs; developed a method and pipeline for KL-based self-distillation.

Publicis Re:Sources Paris

AI / ML Intern

Developed preprocessing methods and a training pipeline for precise placement of objects in images generated by latent diffusion models. Project note.

Candela Technology Stockholm

Summer Intern

Swedbank Stockholm

Client Advisor

Lidingö Jolleseglare Stockholm

Catamaran Coach

Swedish Armed Forces K3 Hussar Regiment

Reconnaissance & Infiltration Specialist

KSSS — Royal Swedish Yacht Club

Sailing Instructor, responsible for the competitive racing group.

KTH Royal Institute of Technology Stockholm

Degree Programme in Engineering Physics, MSc in Applied Mathematics for Data Science.

École Polytechnique (l’X) Palaiseau

Third year of Cycle Ingénieur, track in Applied Mathematics.

Kungsholmens Gymnasium Stockholm

Natural Sciences Program; participated in the Young Entrepreneurship Program.

Tokenmaxxing!

Featured in The New York Times on maximising the use of AI. Read article.

Vocabulary Expansion of LLMs via KL-Based Self-Distillation

Ericsson / KTH master’s thesis. Read on arXiv.

Founders House Fintech + AI Hackathon

Winner, “Most Likely to Become a Unicorn.” Sponsored by OpenAI, EQT, and Stripe. Event note.

Controllable Clothing: Precise Labels and Generation

Publicis / l’X technical report. Read on arXiv.

Marseillestiftelsens Stipend

Awarded by Tessinsällskapet.

Swedish Championship Silver

F18 catamaran class sailboat.

PyTorch, FSDP, DeepSpeed, TRL, PEFT, vLLM, and SGLang.

Slurm, Kubernetes, Docker, AWS, Azure, sandboxing, and MCP servers.

Swedish (native), English (fluent), French (fluent).