PyData London - 74th meetup

Venue: 1 Angel Lane, EC4R 3AB

Please note:

1. A valid photo ID is required by building security. You MUST use your full real names on your meetup profile; otherwise, you will NOT be allowed to enter the building at the night of the event!
2. This event follows the NumFOCUS Code of Conduct; please familiarise yourself with it before the event.

Tickets are assigned through a lottery draw about one week before the event.

If your RSVP status says "You're going." you can get in. No further confirmation is required. You will NOT need to show your RSVP confirmation when signing in.

If you can no longer make it, please unRSVP as soon as you know so we can assign your place to someone on the waiting list.

***

Code of Conduct:
This event follows the NumFOCUS Code of Conduct; please familiarise yourself with it before the event. Please contact the organisers with any questions or concerns regarding the Code of Conduct.

***

As always, there'll be free food & drinks generously provided by our host, Man Group.

***

Main Talks
1️⃣ Billion-Scale Facial Recognition - Sefik Serengil

In this talk, we demonstrate how to perform billion-scale facial recognition using the deepface package for Python and the approximate nearest neighbour algorithm. We have a database of billions of faces, and we show how to create embeddings for each face with deepface and use Faiss to search for an identity in milliseconds, even in such a large database.

2️⃣ How to build a scalable robots ensemble to collect big microscope imaging data - Pavel Katunin and Anton Nikolaev

We are trying to find a method that will allow us to automatically crack protocols of predictable differentiation from stem cells into any needed cell type (f.i. human neurons)
We use microscopy, CV algorithms, optimizing algorithms, machine learning, and robotics for that purpose.
I'm going to speak about the first stage of our project, where we created a cheap open-source microscope that we are using for data collection and manipulating samples and substances.

⚡Lightning Talks ⚡
⚡1️⃣ ⚡ LangChain, GPT-3, and Me: Promises and Perils of Language Models - Jam Coombes

Between GPT-4, Bard and LLaMa.Language Models (LM) have recently been in the news, causing excitement and consternation beyond the AI community. Many of these models are released via APIs or released as open-source, so engineering with these tools is easier than ever.
How, can we, as Pythonistas work with these whimsical, creative and intelligent tools to engineer useful tech?
This informative and thoughtful talk will explore the promises and perils of these language models. The paper "GPTs are GPTs", which talks about OpenAI's idea of how this technology will be used in the future of work. The near-term risks of bias and misinformation are considered.
We also haven't invented the technology yet that keeps a powerful and wilful agent aligned with our values. Possible solutions, like Reinforcement-Learning from Human Feedback and Mechanistic Interpretability are considered, with relevant links to where the open-source community can get stuck in and help out.
The OpenAI API is demonstrated, and the LangChain library for composing language models is demonstrated. This library thinks in terms of LMs, prompts, chains and agents. It uses utility chains and Language model chains. Then, it makes them easy to chain and compose into one another, similar to piping command line utilities into each other on mac and linux.
The LangChain Python DataFrame agent integrated as a Slackbot takes an input CSV and answers natural language questions about it, there are similar open-source agents within the library offering natural language question answering over SQL databases, totally changing the game for Data Teams.

⚡2️⃣⚡ Software Engineer to Machine Learning Engineer - Besart Shyti

I was a software engineer for six years, and then retrained as a machine learning engineer. I'm currently working at Meta on the recommendation engine.
I want to talk briefly about how I made the transition and also let people know that I'm running a free non-profit course for developers who want to start building ML-driven products.

Logistics
Doors open at 6.30 pm (get there early as you have to sign in via building security), talks start at 7 pm, and drinks from 9 pm in the bar. We will have reduced capacity for this event, but there will be plenty of people to discuss data science questions with!

Please unRSVP in good time if you realise you can't make it. We're limited by building security on the number of attendees, so please free up your place for your fellow community members!

Follow @pydatalondon (https://twitter.com/pydatalondon) for updates and early announcements.