Kristian Hartikainen
Helsinki, Mainland Finland
Summary
🤩
RockstarKristian Hartikainen is a Founding Engineer and Head of Software at Eka Robotics with 13 years of experience blending research-grade machine learning with production software engineering. He earned a PhD at the University of Oxford after research stints at DeepMind, Google Brain Robotics, and UC Berkeley’s RAIL lab, contributing to landmark work in model-free RL and maximum-entropy policies. Kristian has shipped cloud-scalable tooling and infra (notably GCP autoscaling integrations for Ray) and improved core ML libraries, adding conditional RealNVP support in TensorFlow Probability and refining MuJoCo and Gym environments. He bridges deep algorithmic insight and pragmatic engineering—equally comfortable implementing MPC tweaks for humanoid control or hardening autoscaler log syncing. An active open-source maintainer, he quietly surfaces research ideas into widely used tooling, accelerating reproducible RL experiments for the community. Based in Helsinki, he pairs academic rigor with startup execution, leading software for a robotics company from day one.
14 years of coding experience
Github Skills (41)
environmental10
docker10
c-language10
dev-environment10
probabilistic-programming10
python10
gymnasium10
testing10
openai-gym10
machine-learning10
cloud-infrastructure10
dockers10
reinforcement-learning10
enviroment10
gcp10
Programming languages (8)
Github contributions (5)
rail-berkeley/softlearning
Apr 2018 - Feb 2021
Softlearning is a reinforcement learning framework for training maximum entropy policies in continuous domains. Includes the official implementation of the Soft Actor-Critic algorithm.
Role in this project:
ML Engineer
ML EngineerContributions:848 commits, 114 PRs, 68 pushes in 2 years 10 months
Contributions summary:Kristian made extensive contributions to adding and integrating Ray support, a distributed computing framework, into the reinforcement learning framework. Their work involved modifying and creating new example scripts, such as `mujoco_all_ray.py` and `multigoal_ray.py`, to leverage Ray's capabilities. The changes included updates to algorithm implementations, environment initialization, and configuration for distributed training. Furthermore, the user implemented improvements to the logging system and contributed to the handling of observations.
actor-criticreinforcement-learningsoft-actor-criticdeep-learningdeep-reinforcement-learning
haarnoja/sac
Nov 2017 - Aug 2018
Soft Actor-Critic
Role in this project:Back-end Developer / ML Engineer
Back-end Developer / ML EngineerContributions:651 commits, 4 PRs, 4 pushes in 8 months
Contributions summary:Kristian primarily contributed to the Soft Actor-Critic (SAC) algorithm implementation. Their work focused on implementing and modifying core components, including adjustments to the Q-function and value function, and implementing a Latent Space Policy. The user integrated and refined the RealNVP bijector within the policy. They also worked on adding new environments.
actor-critic
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.