ClePy July Meetup - PyOhio Preview!

We will be hosted at Happy Dog, in the Underdog (basement).

Meetup Agenda
6:00-6:30pm Social and Setup, Announcements.
6:30-7:30pm Presentations (see below)
7:30-8:00pm Social and Clean-up

We are excited to preview two talks from PyOhio 2026 . This is an opportunity for you to attend these talks before the conference. It is also an opportunity for the speakers to practice their talks and gather feedback.

Talk 1: My Model Works Locally. Why Is Production Lying to Me?
by Dinky Mishra
You tested it. You validated it. It worked perfectly on your machine. Then you deployed it, and everything fell apart.
Production failures are not random. They follow patterns, and once you know them, you will start spotting them before they cost you a week of debugging. Through a structured walkthrough of real failure scenarios, this talk breaks down the five most common reasons ML models and Python applications behave differently in production than they did during experimentation: training-serving skew, environment mismatches, silent type errors, data assumptions that stop being true, and non-determinism from unseeded randomness.
Each failure mode is paired with a clear diagnosis and a concrete fix you can apply before your next deployment.
Who this is for: Anyone who has deployed a Python application or ML model and been surprised by what happened next. Also valuable for anyone preparing to deploy for the first time.

Talk 2: Speeding Up Clinical Trial Analysis with Python
by James Austrow
What do you do when the standard way to solve a problem is just too slow? In any field, comparing pairs of items, whether patients, products, or data points, leads to code that grinds to a halt as the dataset grows. The usual advice? “Just wait longer” or “use a faster language.” But what if you could rewrite the rules?
In this talk, we’ll show how the right algorithm can dramatically speed up a statistical test used in medical research: the win ratio. This method helps doctors and researchers evaluate pairs of patients on complicated criteria, but until now, it’s been painfully slow for large datasets. We’ll break down:

• Why the old way is slow, and why most people accept it
• How re-framing the problem led to a smarter solution
• The results: 20-50x speedups for typical trials, and over 100x for huge ones

No advanced math or statistics background required, just curiosity about how Python can solve real-world problems in unexpected ways.
Takeaways:

• Learn how algorithmic thinking can turn a slow process into a fast one
• See how pure Python can outperform optimized C++ code in some cases
• Get inspired to look for hidden inefficiencies in your own projects

PyOhio 2026 is on July 25-26 at the Cleveland State University Student Center. Register for free here: https://www.pyohio.org/2026/

---
If you decide to not come to the meetup but initially RSVP yes, please change your response so we have a proper headcount and folks who are on the waitlist can attend.

Want to present a talk? Let us know on meetup or the #clepy channel on Cleveland Tech Slack.

Join the Cleveland Tech Slack group here: https://cleveland-tech.vercel.app/