Aarni Koskela
CTO Founder Left Shark at Valohai
Turku, Southwest Finland, Finland
Summary
🤩
Rockstar🎓
Top SchoolAarni Koskela is a seasoned developer-generalist and entrepreneurial CTO with 23 years of experience building across web, desktop, embedded, and low-level systems from Turku, Finland. As founder and CTO of Left Shark (Valohai) he blends hands-on engineering with product leadership, having previously led Anders Inno for nearly a decade. He contributes widely to open source—touching notable projects like pandas, Hugging Face Accelerate, llama.cpp and Babel—where his work often focuses on code quality, testing, performance and cross-language interoperability. Comfortable across Rust, Python, C/C++ and frontend stacks, he frequently refactors for maintainability and enables modern tooling (e.g., Ruff, f-strings, CI improvements). Colleagues know him for pragmatic fixes that improve long-term reliability—tests, docs and small UX touches—and for quietly modernizing legacy codebases without fanfare.
23 years of coding experience
9 years of employment as a software developer
Turun yliopisto - University of Turku
Stackoverflow
160,822reputation
8.7mreached
5,573answers
2questions
boto
top-5%
dictionary
top-5%
regex
top-5%
oop
top-5%
pandas
top-5%
subprocess
top-5%
Github Skills (200)
threejs10
dis10
testing10
command-line-interface10
instrumentation10
service-account10
ubuntu10
hyperparameter-optimization10
encoding10
linter10
quantization10
schema10
code-optimization10
shader10
redis-client10
Programming languages (28)
Github contributions (5)
bitsandbytes-foundation/bitsandbytes
May 2023 - Jan 2026
Accessible large language models via k-bit quantization for PyTorch.
Role in this project:
ML Engineer
ML EngineerContributions:103 reviews, 25 PRs, 233 comments in 2 years 8 months
Contributions summary:Aarni primarily contributed to the `bitsandbytes` library, focusing on fixing typos and improving CUDA support within the PyTorch ecosystem. They addressed minor code errors, improved CUDA support detection in tests, and corrected type annotations. Their contributions also involved updates to the underlying Triton kernels and other modules related to 4-bit quantization.
large-language-modelspytorchquantizationllmmachine-learning
shuup/shuup
Jun 2015 - Feb 2016
E-Commerce Platform
Role in this project:Back-end Developer
Back-end DeveloperContributions:2 releases, 373 commits, 203 PRs in 8 months
Contributions summary:Aarni made significant contributions to the e-commerce platform by updating the Sphinx configuration for documentation, adding documentation for add-ons, and removing a Stripe-related module. They also addressed Python 3 compatibility issues and corrected usages of Category ancestor retrieval within templates. Moreover, the user added contact group administration and made general improvements to order processing.
ecommercedjangopythonecommerce-platformecommerce-shopping-solution
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.