Mårten Saltin

AI engineer building agents end to end.

Hello — I'm Mårten, in Sweden.

Get in touchDownload résumé
About

How I got here.

Programming has always been close to my heart. It is an absolute gold mine for creating significant action in any field, and I want to contribute to it. I studied Computer Science and Software Engineering at Linköping University, and hold a master's from there in AI and machine learning.

Growing up between Stockholm, Mallorca and Palo Alto exposed me to very different kinds of academia and culture. Those years gave me a deeper love for creativity and innovation — the core of exceptional programming — and the habit of adapting quickly to somewhere new.

My passion for programming started young. Today, my expertise spans AI and machine learning, from LLM agents and speech recognition to natural language processing and computer vision. I combine that expertise with a strong foundation in cloud architecture, distributed systems and software engineering to build complete products. My focus is on turning complex AI capabilities into reliable systems that solve practical problems in production.

Outside the academic realm I am passionate about sports and leadership. I spent my high-school years as a part-time soccer coach, mentoring young enthusiasts, running training sessions and managing games and tournaments. My aim is to be an innovative engineer, and through that to make a meaningful impact on the world through the digital space.

Experience

Five years shipping AI into products people use.

Speya

AI EngineerFeb 2026 — Present
  • Built AI sourcing workflows from the agent logic to the chat interface, so users could follow progress live, see which tools were running, stop tasks and pick up earlier sessions.
  • Built the Excel workspace, including an interactive spreadsheet editor, an AI agent for working with Excel, and tools for managing, arranging and formatting workbooks.
  • Connected Gmail and Outlook so users could draft multiple emails and send them from their own accounts. Set up OAuth authentication and tightened access controls for personal mailboxes.
  • Made agents more reliable and efficient with Redis and prompt caching, circuit breakers and fixes based on production traces. Added LangSmith evaluations and resolved bottlenecks in guardrail checks.
  • Reworked the sourcing agent into focused agents and subagents, with reusable skill-based instructions to guide how they handle tasks and choose tools.

VOXO AB

Stockholm, SE

AI EngineerDec 2024 — Feb 2026
  • Fine-tuned an automatic speech recognition model, reducing average Word Error Rate from 0.20 to 0.11.
  • Developed and deployed a retrieval-augmented generation (RAG) system for customer interactions, data analysis, and automated responses.
  • Fine-tuned and benchmarked speaker diarization models to improve speaker recognition.
Junior DeveloperFeb 2022 — Dec 2024
  • Built a scalable AWS data lake using Athena and Glue, audio-data transfer workflows, and a shared Python library for data access and model training.
  • Fine-tuned and evaluated an NLP model on AWS SageMaker to detect and anonymize sensitive personal data; developed a transcription spell corrector with over 90% accuracy.
  • Built an NLP system that analyzed key events and generated presentation-ready insights, reducing more than 20 hours of manual work to seconds.

Augmatica AB

Stockholm, SE

Software DeveloperSummer 2024
  • Developed React components with the team and monitored server performance to support system reliability.

Linköping University

Linköping, SE

M.S. Artificial Intelligence and Machine LearningAug 2024 — June 2026
B.S. Computer Science and Software EngineeringAug 2021 — June 2024
  • Led an eight-student bachelor's thesis team.
Skills

The stack I reach for.

Languages
  • Python
  • TypeScript
  • JavaScript
  • SQL
  • C++
  • C
  • Java
  • R
  • CUDA
  • C#
AI & ML
  • LLM agents
  • LangGraph
  • LangChain
  • LangSmith
  • RAG
  • NLP
  • ASR
  • Deep learning
  • Reinforcement learning
  • Computer vision
Web & infrastructure
  • Nuxt
  • Vue
  • React
  • Strapi
  • Redis
  • Ably
  • OAuth
  • AWS
  • SageMaker
  • Athena
  • Glue
  • Docker
  • Kubernetes
  • Git
Projects & awards

Robots, cars, and one bronze medal.

Humanoid Robot

Sep 2021 — Feb 2023

Developed object detection and OCR with YOLO, expanded the robot's state machine, and modularized infrastructure with Docker Compose.

Self-Driving RC Car

Aug — Dec 2022

Trained a lane-detection model and integrated the car's software components.

Robotics World Cup

Summer 2022

Competed with the Linköping University team in Bangkok, Thailand.

RoboCup 2022

The Linköping University team at RoboCup 2022 in Bangkok.

BronzeTeam result

Contact

Open to interesting problems.

The fastest way to reach me is email.

Emailmartensaltin@gmail.comPhone+46 72 870 73 63RésuméDownload