Thomas Dybdahl Ahle
Copenhagen, Capital Region of Denmark
Summary
🤩
RockstarThomas Ahle is Head of ML and Staff Research Scientist at Normal Computing in Copenhagen with 19 years of experience building efficient, production-ready machine learning systems grounded in rigorous algorithmic research. He bridges theory and practice—publishing optimal LSH and sketching methods in top venues as a postdoc and later leading Facebook’s ML Efficiency group where he halved embedding table sizes, developed sketching-based Bayesian inference, and designed low-memory transformers for mobile deployment. His work favors simple, robust algorithms that translate cleanly into production, with a special interest in resource-constrained and auto-formalizing systems. An active open-source contributor, he built one of the most-used Linux chess clients (PyChess) and has contributed practical optimizations to projects like smhasher and Stanford’s DSPy. He holds a PhD in Computer Science and a First-class BA from Oxford, combining deep theory with hands-on engineering.
19 years of coding experience
Stackoverflow
31,206reputation
9.1mreached
192answers
72questions
sql
top-5%
image
top-1%
hide
top-5%
concatenation
top-5%
java
top-5%
mysql
top-1%
Github Skills (43)
algorithm10
c-programming-language10
algorithms10
c-language10
uml10
python10
mysql10
image10
testing10
datastructure10
machine-learning10
datastructures10
data-structure10
listview10
machinelearning10
Programming languages (15)
Github contributions (5)
stanfordnlp/dspy
Feb 2024 - Dec 2024
DSPy: The framework for programming—not prompting—language models
Role in this project:
Back-end Developer
Back-end DeveloperContributions:22 reviews, 44 PRs, 24 pushes in 10 months
Contributions summary:Thomas primarily focused on developing and refining features within the DSPy framework, as evidenced by the creation of a new Signature class with associated tests and types. Their work also involved updating the load/dump state functionality to utilize the newly created signature, enhancing the framework's internal workings. Furthermore, the user made significant contributions to the test suite, implementing functional tests to ensure the correctness and reliability of the framework's core functionalities.
nlpbertknowledgepredictlanguage-models
rurban/smhasher
Sep 2020 - Sep 2020
Hash function quality and speed tests
Role in this project:Back-end Developer
Back-end DeveloperContributions:1 review, 8 commits, 7 PRs in 7 days
Contributions summary:Thomas primarily contributed to the implementation of hash functions within the `smhasher` repository, as indicated by the commit messages and code changes. They introduced new hash function implementations, specifically focusing on multiply-shift and polynomial hashing techniques. The contributions also involved optimization efforts to improve the performance of the hash functions, particularly handling odd byte numbers, and fixing potential memory issues, demonstrating a focus on speed and correctness.
speedhashsha256hash-functionc-plus-plus
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.