Importance of Community
It's the week after the end of PyCon UK 2025 and I'm reflecting on what a great conference it was. This is the third PyCon I've been to, and the second PyCon UK. This year, I really wanted to push myself to give a talk and submitted a few proposals for consideration. My talk on Synthetic Data was accepted a few months ago, which of course meant I didn't do any more work until a couple of weeks out. Going to the conference, even as an attendee is a great experience, giving a talk or leading a workshop does increase the stress levels but it gives you an opportunity to interact with so many more people and give back to the community. I loved it, and hope to do a few more in 2026.
No data, No Problem
Synthetic Data using LLMs and Faker
My talk, on the face of it, was about generating synthetic data to mimic user journeys through a website, but baked into it was an exploration of when to use AI versus a software engineering to solve a problem. I started first with an LLM, but when the results weren't good enough, I switched to building this tool using Faker, a Python library that can generate fake data for a number of different purposes. You can watch the full video, but here's a summary of my findings if you don't have the time to watch the video:
Use LLMs when:
- The dataset you want to generate is small (< 1000 records)
- Logical consistency isn't critical (for example, if a user has a different name in their email address to what's specified in their name field)
- You need it once and don't plan to repeat the data generation process
- The data structure is simple and easy to describe (even better if you have a schema to describe it)
- You don’t need large amounts of variability in your synthetic data.
Use Faker when:
- You need complex, highly consistent data
- Performance matters and you need to generate larger volumes of data
- You'll be generating data repeatedly
- Multiple team members need to use the tool
- You need precise control over data relationships
Key Learnings
- Natural language has limits: When the desired outcome is hard to describe precisely in natural language, code becomes a better specification tool.
- Consistency is hard: LLMs struggle with maintaining logical relationships across large datasets.
- Reusability matters: Building a configurable system pays dividends when you face the same problem repeatedly.
- Performance scales differently: Faker's deterministic approach scales linearly, while LLM generation can become exponentially slower.
Community in the Age of AI
If you're thinking about attending a conference, stop hesitating and buy that ticket.
They are such a wonderful experience and provide fertile grounds for the sort of random connections you just don't get in our digital lives. In the Age Of AI, it's even more important to show up and connect with humans, because it is all too easy to just stay in our own personal and professional bubbles.
Along with connections, there's opportunities to see how other people are solving problems in their domains. This gives you an opportunity to really expand your toolkit and learn more about your chosen technical field. There's so much serendipity that happens when you turn up to a random talk about Raspberry Pis and Crime or the Luddites and How to Play the LongGame in the age of AI. The energy is infectious too. After three days of learning, I was so inspired that I started coding improvements on my train journey home. That's the conference effect: you leave buzzing with new ideas and the motivation to implement them.
As a stretch goal, why don't you try volunteering or presenting a talk! Both these options are so rewarding and really helped me get even more out of the event. If you use Python, you're local Python community is always looking for volunteers.
Future PyCons
So how do you find out about the next PyCon? These websites maintain a good list of upcoming conferences :
And if you want to put in a proposal this handy website lists how long you have until a proposal deadline for your chosen conference:
👆 I found out about this website through a Lightning Talk I went to at PyCon DE 2025!