Iulia Feroli

Iulia Feroli

@iuliaferoli.com

👩‍💻Tech stuff, conferences, ML/AI 🥑Dev Advocate @Elastic 🔎🫎

104 followers73 following72 posts

Pinned

Intro post: Iulia talks about tech stuff 🐍👩‍💻🔎📊 I make content (blogs, YouTube videos, conference talks, repos) on techie stuff. Mostly data science & ML oriented. I’m an Elastic developer advocate - how to make super cool, semantic search apps. #python #tech #dev #elastic #ml

Just when I finally used up months worth of #audible credits to get all the #HarryPotterAudiobooks, they announce the new full-cast ones are coming out soon 😭 It was hard enough to pick between Stephen Fry and Jim Dale

Gone from filming in every random corner of my student apartments to finally having a fixed "studio" for my videos. Only took 6 years since the first #youtube video 😅

A 300h Civ Six girly just bought Europa Universalis, Anno 1800 and Old World in the 90% off autumn steam sale. Wish me luck, won’t resurface for a few days.

I've been releasing loads of new content on our @elastic.co YouTube community channel. Covered #LLMs, #RAG, #agents, #MCP, and the latest on #contextengineering. Yes, I'm basically playing buzzwords bingo / catch 'em all. And yes, I've also mastered "YouTube thumbnail posing". #hatethegame

I am sooooo beyond excited to keynote #PyCon Sweden this year!! This is a truly special event with a lovely community in a wonderful city. I am so proud to come back in this role and I hope the talk will be useful or at least interesting to the pythonistas joining ;) See you on halloween!! pycon.se

#CDN Romania thank you for an amazing conference experience! Great crowd and discussions, amazing speakers, lovely organisers, good location and food <3 I was so happy my talk about building @elastic.co analytics dashboards from your data (like an accurate DYI Spotify Wrapped!) was well received!

The only way we can get great results out of LLMs is if you perfect prompt engineering / figure out how to be as explicit and precise as possible about your requirements and how the code should perform. Sooo is the AI revolution going to actually make people write good documentation and wikis?? 😅😅

It turns out #PHP is as far away from my familiar safety blanket of #Python as you can get 😅. I am tracking and notes & journey so far (probably soon to be a blog) My "WTH is that" list so far: npm, vue, bun, brew, blade, react, mix, vite, herd, inertia, sail, tailwind, nginx.

Second day at #laracon EU and I have to say, the community here is so lovely and exciting that it made me want to attempt to learn PHP! And that's something since in all my years of programming so far I never once got tempted to touch #javascript. I'll call that a big #laravel win!

Quick dinner in the uni campus canteen: on one hand - awesome healthy cheap food. However - you have to listen to someone loudly ponder if nietzsche’s philosophy is compatible with the values of Christianity; and mansplain the moral of the great Gatsby all within 30 seconds. Gen Z be wildin

At #laracon EU today in Amsterdam with our elastic booth. Come say hi if you’re around! Seeing a lot of great use cases especially with our open telemetry integrations for monitoring and observably w/ the laravel ecosystem

Today I dove into my activity tracker data; specifically, the sleep score that Oura calculates daily. The intuition/assumption about it is pretty simple - more deep sleep and lower heart rate correlate with better scores. Is it that simple? Let's take a look in a quick data science experiment

View full thread

Now to look at what days this gets back, and check what the sleep scores for those days was. Surprise, surprise. It's pretty much the same dates we got in the initial query. This confirms that the intuition/hypothesis was pretty much spot on and I can "predict" my sleep score with simple math

Why does this matter? You can look a simple "sleep score" and believe it's a fancy, groundbreaking AI insight. However, more often than not - it's either a reasonable & intuitive formula or maybe a basic regression. Simpler. More accurate. Cheaper to compute. Choose smart data science over trends!!

Today I dove into my activity tracker data; specifically, the sleep score that Oura calculates daily. The intuition/assumption about it is pretty simple - more deep sleep and lower heart rate correlate with better scores. Is it that simple? Let's take a look in a quick data science experiment

View full thread

Finally, the experiment. Can it be as simple as just making a basic formula around sleeping times and heart rate? I build a query looking at deep sleep over 1.5 hours, heart rate under 60, and sorting by highest REM time:

Today I dove into my activity tracker data; specifically, the sleep score that Oura calculates daily. The intuition/assumption about it is pretty simple - more deep sleep and lower heart rate correlate with better scores. Is it that simple? Let's take a look in a quick data science experiment

View full thread

Step 2: we need a few lines to add the new data into an elasticsearch index so we can look through it easily. This was also super easy - as it tends to be with json files - so we didn't need any additional mapping or data processing steps.

Now comes our tiny data science experiment or hypothesis. I gathered the days with the highest sleep scores with a simple sort on the score field. sort="readiness.score:desc"

Today I dove into my activity tracker data; specifically, the sleep score that Oura calculates daily. The intuition/assumption about it is pretty simple - more deep sleep and lower heart rate correlate with better scores. Is it that simple? Let's take a look in a quick data science experiment

First up I found the Oura developer API and got a quick dump of my activity/sleep data. Oura itself seems to have the most restrictive native way to look at your data (vs fitbit or garmin) so this is already a great find/idea if you want some more insights!

New year new project. I've been seeing lots of comments about the ~accuracy~ of spotify wrapped at the end of 2024 - so I've decided to build my own with #elasticsearch It's actually super easy to put together and you can do way more drilling into your trends/taste over time <3 Blog coming soon!

In the approaching spirit of starting hobbies as New Year’s resolutions I’m really getting into this MIT open course in finance. (Recorded pretty much the week of the 2008 crash! What a fascinating time capsule.) It’s like the content spectrum opposite of TikTok get rich quick schemes.

ocw.mit.eduFinance Theory I | Sloan School of Management | MIT OpenCourseWareThis course introduces the core theory of modern financial economics and financial management, with a focus on capital markets and investments. Topics include functions of capital markets and financia...

Let's keep it positive. First dig into "starter pack" since I think it's such a cool feature on bsky! These are some of the most positive posts about it so far. Curious to get some aggregations on most popular starter pack topics 💭🤔

So far I can see there's about 3500 posts about starter packs out of the 2million entried in the dataset. Which is still half of the almost 9k posts about "follow".

Tried to run a negative sentiment search matching "AI" on the bsky posts dataset to analyze how people feel about it, maybe gather some insights. Main insight: it's time to install some kind of profanity filter 😅

Started playing with huggingface and elasticsearch and I found this really cool dataset of 2M bluesky posts. I can now *factually* confirm that a lot of people just use bluesky to talk about twitter 😅 Starting to look into the sentiment of these posts next. I'll share any interesting insights 😉

Working on the new conference session for the coming tech events cycle. Working titles so far: “We have genAI at home” “Taking the large out of LLMs” “Is it genAI or just a reasonable ML use case?” “GenAI: hitting a nail with a rocket ship” Can you guess where I’m going with this 😅

If this doesn’t fully embody the spirit of “devs will just find a way to make it work” DYI vibes, I don’t know what does 😂 This is my 4k vlogging camera we duct taped to a wall. Aka our top of the line conference recording system. You can always find a way. 🎥🩵