About
I am a machine learning developer focused on developing open-source tools and educating…
Activity
-
If you're a data scientist or coder, this practical book by Lewis Tunstall, Thomas Wolf, and Leandro von Werra shows you how to train and scale these…
If you're a data scientist or coder, this practical book by Lewis Tunstall, Thomas Wolf, and Leandro von Werra shows you how to train and scale these…
Liked by Lewis Tunstall
-
💬🔥Releasing idefics2-8b-chatty, the chat-optimized version of Idefics2! It is a very efficient (8B parameters) state-of-the-art VLM, has been…
💬🔥Releasing idefics2-8b-chatty, the chat-optimized version of Idefics2! It is a very efficient (8B parameters) state-of-the-art VLM, has been…
Liked by Lewis Tunstall
-
Is there any research on which prompts produce an LLM-judge that is most correlated with human preferences? I'm aware of the canonical work by LMSYS…
Is there any research on which prompts produce an LLM-judge that is most correlated with human preferences? I'm aware of the canonical work by LMSYS…
Shared by Lewis Tunstall
Experience
Education
Licenses & Certifications
-
Getting Started with Stream Processing Using Apache Flink
Pluralsight
-
Understanding Apache Flink
Pluralsight
-
Structuring Machine Learning Projects
Coursera
Publications
-
Unsupervised Anomaly Detection for Seasonal Time Series
IEEE
We extend eBay's Atlas algorithm to automatically detect anomalies in unlabeled, seasonal time series data. Named MULDER, the algorithm involves deriving a "surprise" metric from the time series, which is then analysed statistically for anomalies. We evaluate the efficacy of MULDER via the Numenta Anomaly Benchmark, and calibrate it for deployment with injected anomalies on production data. We find that MULDER can be used to create alerts with a low false positive rate, and outperforms several…
We extend eBay's Atlas algorithm to automatically detect anomalies in unlabeled, seasonal time series data. Named MULDER, the algorithm involves deriving a "surprise" metric from the time series, which is then analysed statistically for anomalies. We evaluate the efficacy of MULDER via the Numenta Anomaly Benchmark, and calibrate it for deployment with injected anomalies on production data. We find that MULDER can be used to create alerts with a low false positive rate, and outperforms several popular open source implementations.
Other authorsSee publication -
Light stops, blind spots, and isospin violation in the MSSM
Journal of High Energy Physics
-
ΔI = 1/2 rule for kaon decays derived from QCD infrared fixed point
https://journals.aps.org/prd/abstract/10.1103/PhysRevD.91.034016
Honors & Awards
-
Swiss Government Excellence Scholarship
Federal Commission for Scholarships for Foreign Students
-
Dean’s Commendation for Doctoral Thesis Excellence
The University of Adelaide
-
Fulbright Scholar
Australian-American Fulbright Commission
-
H.S. Green Prize
The University of Adelaide
Languages
-
English
Native or bilingual proficiency
-
Spanish
Professional working proficiency
-
German
Limited working proficiency
More activity by Lewis
Want to dig into large-scale AI training infrastructure? 🧠 I will present how we set up and manage large H100 GPU clusters for LLM training at…
Liked by Lewis Tunstall
💾 LLM Datasets LLM development is increasingly moving towards curating high-quality datasets, as shown by Llama 3. I've compiled a collection of…
Liked by Lewis Tunstall
Direct Nash Optimization (DNO) better than DPO for RLHF/RLAIF? 🤔 DNO is a new RLHF method from Microsoft Research using a batched on-policy…
Liked by Lewis Tunstall
🦙 Can we use Llama 3 to fine-tune smaller models? ⚗️Model distillation and synthetic data are a promising topic. But every single method…
Liked by Lewis Tunstall
🆕 Which RL agent is the best? 🏆 It's now possible to find out, thanks to the Open RL Leaderboard ! ➡️ https://lnkd.in/dadaianz 🧩 Features: -…
Liked by Lewis Tunstall
Lewis Tunstall 's slides about how to align LLMs to human preferences! Great, tnx! https://lnkd.in/dGQbiJ62 #LLM #RLHF #DPO #KTO #ORPO
Liked by Lewis Tunstall
People also viewed
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore MoreOthers named Lewis Tunstall
-
Lewis Tunstall
Student at University of Sussex
-
Lewis Tunstall
--
-
Lewis Tunstall
--
-
Lewis Tunstall
Chef looking for a career change
5 others named Lewis Tunstall are on LinkedIn
See others named Lewis Tunstall