Daniel Gutierrez
banner
ddgutierrez.bsky.social
Daniel Gutierrez
@ddgutierrez.bsky.social
Data scientist, Principal AI Industry Analyst and Influencer - http://radicaldatascience.ai, having a blast teaching data science at UCLA #AI #DataScience #MachineLearning #Rstats
Most ML failures aren’t technical. They’re:
— Versioning issues
— Poor stakeholder alignment
— Lack of monitoring
Data science ≠ Kaggle. It’s shipping, integrating, maintaining.

#DataScience #MachineLearning #AI #RStats
November 10, 2025 at 5:23 PM
aiOla unveils Drax, an open-source speech model with state-of-the-art accuracy and up to 5× faster than models from OpenAI and Alibaba -- radicaldatascience.wordpress.com/2025/11/07/a...

#AI #OpenSource
aiOla unveils Drax, an open-source speech model with state-of-the-art accuracy and up to 5× faster than models from OpenAI and Alibaba
aiOla, a voice-AI lab advancing speech recognition technology, is announcing Drax, an open-source AI model that brings flow-matching-based generative methods to speech. By rethinking how speech mod…
radicaldatascience.wordpress.com
November 7, 2025 at 5:29 PM
LLMs don’t “understand” text.
They predict the next token.
The illusion of fluency can conceal failure modes:
— Calibration
— Attribution
— Security
Understanding how they work is critical to knowing when they fail.

#AI #GenAI #LLM
November 7, 2025 at 5:18 PM
What’s one model, concept, or technique you once struggled with—until one key example unlocked it?
For me: — PCA (until I mapped it to variance in rotation)

#DataScience #MachineLearning #RStats #SignalOverNoise
November 6, 2025 at 7:07 PM
I’ve taught dozens of intro data science courses.
The hardest concept isn’t regression or classification.
It’s thinking conditionally.
— "Given X, then Y."
— "Controlling for Z…"
Statistical reasoning is harder than syntax.

#DataScience #MachineLearning #Statistics #AI
November 5, 2025 at 11:41 PM
The GenAI gold rush is shifting. It’s no longer about training frontier models—it's about:
— Specialized inference stacks
— Efficient fine-tuning
— Enterprise guardrails
Winning isn’t scale alone. It’s control.

#AI #LLM #GenAI
November 4, 2025 at 7:25 PM
R is still the most productive environment for exploratory data analysis.
— Fast loading
— Strong stat packages
— Concise, readable syntax
For slicing truth from noise in tabular data, #RStats remains unmatched.

#AI #DataScience #MachineLearning
November 3, 2025 at 6:44 PM
While editing the 2nd edition of my book, I kept trimming code and adding more explanation. Good code is useful. But understanding why it works—and when it fails—is what creates good data scientists.

#SignalOverNoise #DataScience #MachineLearning #RStats
October 31, 2025 at 6:23 PM
App Orchid Launches New Conversational Analytics Agent With Continuous Semantic Enrichment for Benchmark-Breaking Accuracy -- radicaldatascience.wordpress.com/2025/10/30/a...

#AI
App Orchid Launches New Conversational Analytics Agent With Continuous Semantic Enrichment for Benchmark-Breaking Accuracy
Built on App Orchid’s semantic knowledge graph, the Agent continuously learns from context to improve accuracy, transparency, and enterprise trust. App Orchid, a leader in making data actionable, i…
radicaldatascience.wordpress.com
October 30, 2025 at 5:42 PM
The scariest thing about GenAI isn’t the hallucination rate. It’s the overconfidence in the outputs. A wrong answer said fluently is more dangerous than silence.
Trust layers and human-in-the-loop systems are now table stakes.

#AI #LLM #GenAI
October 30, 2025 at 4:42 PM
BREAKING TODAY! Token Security Launches Token Research to Deliver Actionable Insights into Agentic AI Security Risks -- radicaldatascience.wordpress.com/2025/10/29/t...

#AI #Security
Token Security Launches Token Research to Deliver Actionable Insights into Agentic AI Security Risks
Unit will help protect organizations against emerging threats with  vulnerability analysis, threat intelligence, and remediation guidance Token Security, a leader in Agentic AI and Non-Human Identi…
radicaldatascience.wordpress.com
October 29, 2025 at 7:19 PM
What's one tool you love in R but struggle to replicate in Python? For me: ggplot2 still sets the gold standard for layered, expressive graphics. Even after years of using matplotlib + seaborn.

#RStats #DataViz #DataScience #MachineLearning
October 29, 2025 at 6:35 PM
Everyone’s watching LLM capabilities. But the real action? AI infrastructure. Foundry tools, vector DBs, fine-tuning stacks, and orchestration layers will define the next phase. The model is just the engine. The system wins the race.

#AI #LLM #GenAI
October 28, 2025 at 7:24 PM
When teaching regression, I always emphasize this:
The model is only as honest as your assumptions.
Check:
— Linearity
— Independence
— Homoscedasticity
— Normality
Violating these? You’re interpreting fiction. #RStats

#DataScience #MachineLearning #AI
October 27, 2025 at 9:58 PM
GenAI isn't replacing data scientists. It’s exposing which parts of our workflows were mechanical all along. Your future value? Critical thinking, domain insight, and narrative translation of data into decisions. That’s not automatable.

#AI #LLm #GenAI #DataScience
October 24, 2025 at 6:19 PM
GREAT to walk into my classroom to see all the copies of my book's recent new 2nd Edition in so many of my student's hands! -- amazon.com/Machine-Lear...

#DataScience #MachineLearning #RStats
@uclaextension.bsky.social
Machine Learning and Data Science, 2nd Edition: An Introduction to Statistical Learning Methods with R
This second edition of offers an accessible, hands-on introduction to the core principles of machine learning, statistical modeling, and practical data science—without overwhelming readers with com...
amazon.com
October 23, 2025 at 6:59 PM