Gabe Ketron
banner
gketron.bsky.social
Gabe Ketron
@gketron.bsky.social
Associate Data Scientist @ Cedars Sinai

Opinions are my own.
Dialectical materialism also happens to be a highly effective way to conduct pca dimensional reduction
August 19, 2025 at 8:22 AM
A huge thank you to Dr. Gregory Brewer, Dr. Felix Grun, my co-authors, the UCI MIND Institute, and everyone who supported this work!

🔗 Read more: alzres.biomedcentral.com/articles/10....
Pharmacokinetic and pharmacodynamic assessment of oral nicotinamide in the NEAT clinical trial for early Alzheimer’s disease - Alzheimer's Research & Therapy
Background Nicotinamide, a form of B3 vitamin, is an NAD+ precursor that reduces pTau231 levels via histone deacetylase inhibition in murine models of Alzheimer’s disease (AD). A recent phase 2a rando...
alzres.biomedcentral.com
March 12, 2025 at 12:00 AM
Our findings suggest that oral nic. reached the blood plasma and CSF, however we also found extensive metabolic inactivation to methyl-nicotinamide limiting positive outcomes. Bioavailability and rapid metabolic methylation needs addressing if nicotinamide is further developed.
March 12, 2025 at 12:00 AM
We used mass spectroscopy to measure nicotinamide and its inactive metabolite 1-methyl-nicotinamide in a post hoc, blinded analysis of a randomized clinical trial testing the effect of 1500 mg oral nicotinamide on early signifiers of Alzheimer’s Disease.
March 12, 2025 at 12:00 AM
Once you save the config, and your model is running in your terminal, you can ask your local model for improvements to your existing code, advice on system design in the chat box and even ask it to write code for you. Happy coding!

5/5
January 24, 2025 at 7:36 PM
Once installed, go to the config file and insert the following text to the models section.
"models": [
{ "title": "Llama3.1 8B",
"provider": "ollama",
"model": "llama3.1:8b" }
]
4/5
January 24, 2025 at 7:36 PM
The command will install your model, and allow you to submit questions on the command line. We still need to connect this to your IDE to get code pairing started. Lastly, go to VSCode and install the “Continue” extension, which will allow you to connect your local model.
3/5
January 24, 2025 at 7:36 PM
First, go to ollama.com and install their application. This will allow you to run open-source models with little hassle. Then, go to your terminal and run the command “ollama run llama3.1:8b” .This model is small enough to run on M1 MacBooks without any issues.
2/5
Ollama
Get up and running with large language models.
ollama.com
January 24, 2025 at 7:36 PM
Integrating both LLMs and supportive social structures are critical to boosting research productivity. Balancing tech and human factors are key to unlocking huge leaps in research impact in your research outfit. www.sciencedirect.com/science/arti...
What makes a productive Ph.D. student?
This paper investigates how the social environment to which a Ph.D. student is exposed during her training relates to her scientific productivity. We …
www.sciencedirect.com
January 21, 2025 at 11:04 PM
How can we leverage AIs to ramp up research productivity? Studies show that Ph.D. students produce the most papers and citations when they have highly productive female, mid-career supervisors, and are in small gender-mixed groups of their freshman peers.
January 21, 2025 at 11:04 PM
LLMs were found to significantly boost productivity in complex tasks—consultants using AI completed 12.2% more tasks and completed tasks 25% faster than consultants without AI tools. www.oneusefulthing.org/p/centaurs-a... 📈 #AIProductivity
Centaurs and Cyborgs on the Jagged Frontier
I think we have an answer on whether AIs will reshape work....
www.oneusefulthing.org
January 21, 2025 at 11:04 PM
Fisher–Pitman principles leverage all possible data re-orderings to assess differences. 🔄 By sampling these possible data distributions in a Monte Carlo experiment, tests for small, non-normally distributed datasets become possible. #Statistics
January 21, 2025 at 10:42 PM
The partial derivative structure is set up in such a way that each variate can be calculated in parallel and without the need for caching previous evaluations (see above photo). I'm still a little fuzzy on the math behind this so give it a look yourself! arxiv.org/pdf/2412.00088
arxiv.org
January 13, 2025 at 8:36 PM
To scale high order partial differential equations, the authors use a Taylor expansion of univariate changes to the composition of analytical functions. The Jacobin matrix of this expansion is then partially derived to the order of the number of variates used.
January 13, 2025 at 8:36 PM