Pushing the edge of LLMs and supporting our leadership growth
Good grief, I last wrote in January. How the year can fly past when you've got a bunch of fun projects on the go and you're under the press of overload!
I have notes below on my 1 day deep dive LLM days, an upcoming Success course, growth in my RebelAI leadership group (with some lessons shared) including a London pub invite and finally a note on the upcoming Python By The Sea conference.
I help data science teams that need strategic support to make their projects more valuable, reply if we should chat.
I've been semi-regularly posting on LinkedIn, if we're not Connected, hit me up and maybe note that you read this newsletter so I know to Accept.
I note above that you should UNSUBSCRIBE if this isn't interesting - I haven't written for a while, I won't be upset if you drop off, just hit that link and you'll be cleanly off.
playgroup 1 day LLM deep dive
Just before our PyDataLondon 2025 Conference I ran a 1 day deep dive into LLMs for 13 of my smartest colleagues. We spent the day digging into a kaggle problem that I've been working on for a year, playing with my solution and coming up with lots of alternative approaches.
Half the goal is to look into 'the edge' of what we can do with an LLM today. The other half is to share hard-won stories and show-n-tells around what actually works with GenAI process and tooling.
I put this together because our GenAI world is moving so fast and sensibly addressing leadership's demands to 'do more AI' is rather hard. I figured I wanted a smart posse around me to help share stories on 'this is what actually works, today'.
Using a fully catered training venue we collaborated, talked, laughed and played all day, ran a set of show-n-tells, listed questions from our office, sought answers and ended up in the pub. There's a longer write-up here with a couple of photos.
I'm running another on Friday September 26th (details here), if you're curious there's just a few tickets left. You need to have been using LLMs for over a year, more than just in a copilot situation (e.g. building a data product around it, or equivalent with ML). I'm after practitioners who have hard won experience to share, collaborating with smart peers. If that's you, feel free to reply to this email and ask a question.
I was (quite rightly) called out for the lack of diversity in that audience, that was due to hasty organisation where none of the non-men I know had availability. That's been sensibly addressed for the event at the end of this month and I'll continue to address that for future playgroups.
Successful Data Science Projects course
Some of you will have been on my Successful Data Science Projects course over the last 8 years, this is the only edition I'm running this year, it'll be held virtually on Friday Oct 3rd.
It is aimed at new leaders, anyone who has tried and maybe failed to run successful data science projects. We explore ways our projects commonly fail (mostly - through project setup and people, not because we don't know the tech) and dig into lots of battle tested ways to get to success.
It is a collaborative one day session. Attendees get to know other leaders they can connect with for post-event support. If you have questions, just reply to this email.
Practices from my private RebelAI data science leadership group and a pub invite
Two years back I created my private RebelAI data science leadership group, we've grown to 35 members who meet monthly. We provide a safe space for leadership discussion in our fast moving data science space. If you wonder if you'll fit, if I say "just because we're good at numbers, it doesn't necessarily mean we're good at people" see how your gut reacts.
If a collaborative community could help you make better leadership decisions, write back to me.
This Tuesday 9th in central London I'm hosting our semi-regular pub meet, non-members are invited if they're in a leadership capacity. Building a virtual group (we extend through Europe and meet through Zoom & Slack) is tough, we cement relationships for those in London with these additional meetings. Write back if you're curious.
A little while back I wrote up thoughts on recurring practices including regular questions in Slack on a weekly schedule, our no-agenda coffee calls (two a month), our agenda-based member discussion calls (one a month) and a pub meet every couple of months. These practices help us check in, get to know each other and dig into opportunities and challenges at different tempos around our busy lives.
The regular slack questions include 'how are we doing this month?' where some members are doing great and some are overloaded (and maybe we get to step in and offer some support or a sympathetic ear) and 'what success have we shouted about at work this month?' as a reminder that data scientists probably forget the value of clearly communicating value to colleagues.
I mention the above because if you're in a leadership role, you're not terribly well supported and you lack these sorts of regular interventions that help you look up from the next-task-ahead, maybe you want to think about building yourself a support network (or talk to me about RebelAI, whatever works for you).
In other news
Python by the Sea
My friend Steve Holden is organising a new UK Python conference called "Python by the Sea", near to Hastings on Sept 23rd-25th, the tickets are very reasonably priced and the event is described here. It is an unconference, so the schedule will be organised by attendees, a la Bar Camps (from back in the day).
Catching up
What have I been up to since January? Mostly overload - a big building project on our house ate a chunk of the year. With friends we drove (and won) the motoscape European charity car drive for the 3rd year running. Micha and I got the 3rd edition of our High Performance Python book out. I've spent the year having fun working in private equity as head of data science at Lantern.AI. More on these in future posts.