AI Developer Advocacy & Data Science | The Cool Data Projects Show | Data Moves Me, LLC | LinkedIn Top Voice

"Susie's School Bus Solution", a whimsical rhyming story about the school bus detector that I built for my home using computer vision is now live! From data gathering to deploying for ages 6-12: https://a.co/d/higPAuW

Co-author of "Machine Learning Upgrade: A Data Scientist's Guide to MLOPs, LLMs, and ML Infrastructure" with Wiley. It's available now: https://a.co/d/95jx7QP

I've been a data scientist since finishing my MS in Applied Statistics in 2010. It all started with building neural nets and ARIMAs in the utility industry, then I moved to building models for winback campaigns in healthcare, but the majority of my data wrangling, model building, and hypothesis testing has been in marketing analytics in e-commerce.

The most important thing I learned is that customers sometimes do things that make no sense, so don’t make assumptions, use data.

I’ve shared a lot about ML projects I’ve built for both industry and personal projects at datamovesme.com


New Year, New YOLO 🚍

Ultralytics officially launched YOLO26 on January 14th, and after updating my bus detector, the first thing I noticed for my own use case test images was fewer false positives. 🙌

My system is a narrow-domain CV setup, trained on thousands of images of the same school bus that drives by my house, so latency was never my bottleneck. But YOLO26 still stood out immediately in ways that matter in production. So I’m excited to get to partner with Ultralytics to bring you this post.

YOLO26 moves to a fully end-to-end, NMS-free architecture, which means:

- Fewer post-processing steps
- Cleaner, simpler deployment pipelines
- Less custom code to maintain (especially on CPU and edge devices)

Where YOLO26 shines is in removing work after the model runs. Some highlights that stood out:

- Up to 43% faster CPU inference (workload-dependent, but huge for edge and cost-sensitive deployments)
- No NMS, no DFL → fewer failure modes and less hardware friction
- Better handling of large objects (great for school buses) and small, dense objects

Clearly designed for production systems, not just benchmark wins.

What this really translates to:
- Fewer deployment surprises
- Less post-processing code
- More predictable behavior
- Lower infra costs
- Models that actually ship

If you’ve built detection systems before, you know how much time gets lost after the model runs. 

YOLO26 removes a lot of that complexity, and that’s where the real win is. My use case might be fun, but this is built for real-world impact and deployment.

Try YOLO26 on the Ultralytics Platform:
https://lnkd.in/eh7Wzaxq

And I’m just going to conveniently place the docs right here:
https://lnkd.in/eccT9U-E

Kristen Kehrer


I post blog articles about machine learning projects I complete for work, how to create effective data presentations, data science trends, and about products I believe would help my audience of data professionals and decision makers.

The most important thing I have learned throughout my 12 years of experience as an analytics professional is that customers sometimes do things that make no sense, so don’t make assumptions, use data. Consumer behavior is fascinating.  

Everyone loves to wear their modeling projects like a badge of honor, I’m certainly no different.  Most of my modeling projects have been in e-commerce, while working for companies whose products supported SMBs.  Things like cluster analysis, retention modeling, and I’m most proud of an off-label use case of the TBATS algorithm to determine customers with a seasonal usage pattern, so that we wouldn’t assume the customer’s quiet season was a retention risk.  The first 2 years of my career, as a wee data scientist, I built 100’s of econometric ARIMA models and neural net models in the Utility industry for electric and gas load forecasting.  

Passions include diversity in data and bridging the communication gap between data scientists and the business.

Website:  https://datamovesme.com
Email: kristen at datamovesme dot com

Co-author of "Mothers of Data Science" with Kate Strachnyi

Ultralytics YOLO26 computer vision model

Mavens of Data Podcast Host, [in]structor, Co-Author of Machine Learning Upgrade

Get ready for the first Mavens of Data episode of the new year! 🙌 

Brent Dykes has such deep knowledge and a passion for data storytelling. Come hear him break down some awesome, important, and foundational topics like:

 - What actually qualifies as an insight
 - How to distinguish reporting ("the what") from storytelling ("the why")
 - Why answering “so what?” is the key to influence
 - The role AI will play in generating insights and telling stories

If you've ever been asked for insights and sat there wondering what "qualifies", I hope to see you there!

When: Tuesday, January 20th, at Noon EST
Register: https://lnkd.in/eWWqE7Kj

Let’s get ready to rumble!

There’s no shortage of AI hype on LinkedIn. There is a shortage of people who’ve actually built and operated data systems at scale. That’s why I’m excited about The Great Data Debate 2026.

This is a real debate (and free to watch) between people who’ve built, broken, and rebuilt the data stack across multiple eras.

They’ll tackle the questions many of us are already wrestling with:

 - Will AI analysts actually kill dashboards in 2026?
 - Semantics vs. context vs. ontologies — what really works at scale?
 - Who owns context now: data teams or AI teams?
 - Is context engineering the next evolution of data engineering?
 - What happens to analysts when insight becomes autonomous?

The speaker lineup spans every generation of the data stack, from enterprise platforms to modern analytics to AI-native systems. Are you excited yet?
You’ll hear from:

Bob Muglia (Snowflake)
Tristan Handy (dbt Labs)
Cindi Howson (ThoughtSpot)
Tony Gentilcore (Glean)
Barry McCardel (Hex)
Karthik Ravindran (Microsoft)
Jaya Gupta (Foundation Capital)
Chris Child (Snowflake)

The real shift isn’t in the tools, it’s in how context gets created, owned, and used.

When: Feb 5, 2026 at 11 AM ET

Where: https://lnkd.in/eZsWW7wp 

This might be the most interesting conversation of 2026. Save your spot and come ready to challenge your assumptions.

There's almost nothing I love more than case studies. Tell me how a problem is solved, and I'm going to try and apply that to my use case. The huge bonus is that I can use those learnings again and again until I learn some updated method or technique.

This is why I'm extra excited for the Mavens of Data episode this week with Howard Friedman, Entrepreneur, Author of “Winning with Data Science”,  and long-time practitioner. Howard has so much lived experience in data and I can't wait for him to walk us through the following examples:

1. Site Selection Modeling That Eliminated Costly Mistakes -
How one company improved its revenue forecasts and stopped expensive cannibalization across its retail locations. Howard will walk through what went wrong, how the right product requirements were defined, and what factors were the turning point in improving the forecasts.

2. Loyalty Program Analytics That Drove Double-Digit Growth -
1M+ loyalty members, but the organization wasn’t sure how to monetize them. Learn how cross-functional collaboration and analysis led to targeted offers, higher engagement, and the reactivation of 50,000+ dormant accounts.

3. LLM Automation in Healthcare - Cutting Documentation Time in Half -
Machine-learning use case at one healthcare company reduced documentation time by 2.5X while maintaining 97% accuracy, freeing clinicians to focus on patients, not filling out forms.

Join us :)
When: Thursday, January 29, 2026 @ 11 AM ET
Where: https://bit.ly/4sHsZKE

taplio

Kristen Kehrer's Linkedin Analytics

Kristen Kehrer

Performance ScoreKristen Kehrer's LinkedIn Performance

What is Kristen talking about?

Kristen Kehrer's Best Posts (last 30 days)

Ready to grow your LinkedIn presence?

taplio

.css-6ytdur{background-image:linear-gradient(284deg, rgba(39, 43, 72, 1) 0%, rgba(33, 47, 183, 1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Kristen Kehrer's Linkedin Analytics

Kristen Kehrer

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Kristen Kehrer's LinkedIn Performance

What is Kristen talking about?

Kristen Kehrer's Best Posts .css-1emoo84{font-size:var(--chakra-fontSizes-md);font-weight:400;color:var(--chakra-colors-gray-600);}(last 30 days)

Ready to grow your LinkedIn presence?

Kristen Kehrer's Linkedin Analytics

Performance ScoreKristen Kehrer's LinkedIn Performance

Kristen Kehrer's Best Posts (last 30 days)