Skip to main content

My Latest project using Gen AI

So recently parkrun removed all their stats and as a keen running who is trying to work their way up the top 100 of their local parkrun I wanted to get some of these stats back and have a bit of "fun" at the same time. So here is a little "ETL" process that I developed with the help of Gen AI. 

The steps of my ETL: 

  1. Copy and paste data into Google Sheets template where an AI produced formula extracts URLS from the text and puts them into a new field. This effectively allows me to extract the parkrun athlete id, the primary key, and use it in my analysis. I also have a column to autofill the data I am processing. 
  2. Use an Gen AI generated Google Apps script to process it into a processed sheet, this allows me to build up a backlog of events (I had over 500 to process). 
  3. This is then queried using a Gen AI Google sheets query to extract key information and columns / format times etc.
  4. I then ingest the fully processed sheet into Keboola directly from Google Sheets. 
  5. Within the Snowflake Keboola environment I then create views to summarise and process the data. 
  6. Present the data in Streamlit.  
The above is all saved into a GitHub repo and presented through Streamlit so anyone from Eastbourne parkrun can take a look. It still needs some work and I would love to do some more analysis for individual users. 

Labels:

Comments

Post a Comment

Popular posts from this blog

Gen AI news 29-04-2024

Here are some recent updates and insights related to Generative AI (gen AI) : Enterprise Hits and Misses - Robotics and Gen AI Converge : This article discusses the convergence of robotics and generative AI. It explores breakthroughs needed in the field, the FTC’s policy change regarding non-competes, and the impact on AI model sizes for enterprises 1 . Read more All You Need To Know About The Upcoming AI-Powered OLED iPad Pro : This piece provides a summary of rumors surrounding the next-gen AI-fused OLED iPad Pro, powered by the new Apple M4 chip 2 . Read more Delivering on the Promise of Gen AI : New Electronics reflects on NVIDIA GTC and key announcements that contribute to delivering on the promises made for generative AI 3 . Read more The Future of Generative AI - An Early View in 15 Charts (McKinsey): Since the release of ChatGPT in November 2022, generative AI has been making headlines. McKinsey research estimates that gen AI features could add up to $4.4 trillion to the globa...
Post a Comment
Read more

Keboola Flows

Really finding Keboola was the thing that kickstarted this project otherwise I would be trying to build custom code on a python cloud server and building everything from scratch.  In Keboola you build you data sources and destinations using connection details which is fairly simple and something I will likely cover in another post, same goes for transformations etc. Here though I am going to discuss Flows, this is where you bring everything together. On my free account there are some limitations.  My easiest flow is very basic:  Pull parkrun results e-mail from Gmail to Google Sheets (actually done by Zap not Keboola).  Keboola will, as often as I like, in this case once a week, pull the data from the sheet into its storage.  It will then transfer this to the target database. Currently I have this setup to be MySQL database but I can and might expand that to the Snowflake instance within Keboola.  I then, outside of Keboola, connect to the MySQL database f...
Post a Comment
Read more

Snowflake Scripting - SQL Cursors

Snowflake scripting in SQL seems to be in preview and I have decided to have a play with it. Given how new it is there is limited documentation so I am using a combination of what I can find on the Snowflake site and the odd blog that has been written about it. There appear to be a few quirks, at least when compared to Oracle PL/SQL (though that has been round for years). How many of these are intentional and how many are things to be ironed out I don't know. You can see the procedure I have created it:  Accepts an id as a parameter  Creates a result set selecting from a table, using the parameter as a filter Loads the results set into a cursor.  Loops through the cursor loading the id in the cursor into variable Calls procedure passing in the variable as the parameter.  Then as a proof of concept I tried the Snowflake feature of allowing declaration of variables within the main start and end block. 
Post a Comment
Read more