Vusal

@mdancho84 Does it handle pdf images well?

Microsoft is making moves again. A quiet little Python tool just shot to the top of GitHub’s trending charts. 100,000+ stars. It’s called MarkItDown. And it does something deceptively simple: It turns almost any file into clean Markdown. PDFs. Word docs. PowerPoints. Excel files. Images. Drop a file in. Get structured Markdown out. Sounds small. It’s not. Because one of the biggest bottlenecks in AI workflows — especially RAG systems — is getting messy, real-world documents into a format models can actually use. And real-world documents are brutal. PDFs are chaotic. Word docs are full of hidden formatting junk. PowerPoints are messy and often image-heavy. Spreadsheets can be a nightmare to parse cleanly. That’s where this gets interesting. MarkItDown strips away the friction and gives you something LLM pipelines can actually work with. In other words: less preprocessing, less pain, faster AI implementation. Even better, this isn’t some random side project. It’s an official Microsoft open-source tool. Free. Commercially usable. Practical. I tested it on a 200-page PDF. A few seconds later, I had Markdown that was shockingly clean. And that’s what big tech does at its best: They take an annoying, universal problem that everyone has been duct-taping together… and turn it into a simple standard. That’s why this matters. It’s not just a file conversion tool. It’s infrastructure for the next wave of AI applications. Get it here: github.com/microsoft/mark… 🚨 Want to learn how to build + ship AI and Data Science projects (that businesses actually want)? On April 29th, I am hosting a free workshop to help you get started with AI + DS projects in Python. Register here (500 seats): learn.business-science.io/registration-a…

@MrPassive_ Course

Anyone can buy a Micromarket for $500 down & start making $5,000+/month from it. (One of mine makes $22,000+/month) Today, I’m giving away my Full Micromarket Course for free. All you have to do: • Like this • Comment “Course” & I’ll DM it to you. *Must Be Following Me*

@benkellyone Freedom

Screw it. I want to pay it forward. I’m giving away my *exclusive* guide on how I built my $7M/year business portfolio. • Like this post • Comment “Freedom” And I’ll send you the link. (Must follow, 24 hours only)

@MrPassive_ Course

2025 is the year you place your first Micromarket. (One of mine makes $22,000+/month) Today, I'm giving away my Full Micromarket Course for free. My entire system. • Like this • Comment "Course" & I'll DM it to you. *Must Be Following, 24 Hours Only*

@JinglaiHe Wow!

Something absolutely unbelievable just happened moments ago.. I was in the Library studying and all of a sudden, I got a phone call from Pierre Poilievre (YES YOU READ THAT RIGHT THE PIERRE POILIEVRE). He thanked me for all of the work I put in during the election and told me to keep doing what I'm doing. We ended up talking for over 13 minutes and let me tell you, this man is laser-focused on winning the upcoming by-election and reenergizing the Conservative movement. I only got actively involved in politics when I first started canvassing with my Federal Conservative Candidate over a year ago and never did I ever imagine my political idol would one day recognise the work that I've done and give me a call like he just did minutes ago. Hard work and perseverance does eventually pay off and I'm still in total shock and joy that I had a chance to speak with Pierre. @PierrePoilievre, if by any chance you're reading this, thank you so much again for the call and please know that I along with many of my friends will be working overtime to get you elected as the next PM of 🇨🇦 when the next election happens.

@acoyne You’re so out of touch! Job market has effectively collapsed, and a lot of people are having anxiety of losing jobs.

He's officially panicking. This is awful. Poilievre calls for ‘severe limits’ on Canadian population growth - National | Globalnews.ca globalnews.ca/news/11234497/…

@mcuban Thanks Mark. I’d also add improving healthcare (e.g hospitals, clinics) operations and drug formulary design using AI, specifically operations research and machine learning techniques. I’ve been doing research on this front and administrative cost savings is quite substantial.

Well said. Thanks !

@tm1515152005 @mcuban @LockheedMartin @lockheed Pharmacy Benefit Manager

@mcuban @LockheedMartin @lockheed What is PBM?

Sources in the pharmaceutical industry tell me @LockheedMartin signed an awful PBM contract. So bad that it doesn’t even reimburse the independent pharmacies their employees use at their full cost. That’s just wrong. If you work at @lockheed you should let your CEO know that it’s time to change PBMs

@CTVToronto Fantastic news!

Ontario expanding number of private clinics that can perform OHIP surgeries. toronto.ctvnews.ca/ontario-expand…

@TheZachMueller @huggingface Will there be peft (e.g, prompt tuning) notebook examples as well?

I've created a small little knowledge repository on @huggingface transformers here: github.com/muellerzr/mini… Essentially these contain all the `task` notebooks converted as scripts, showcasing end-to-end usage in under 150 lines of code (but still readable!)

@PhDemetri Would Weibull distribution work too?

New post about estimating the shifted beta geometric distribution for retention analysis Its quick and to the point. dpananos.github.io/posts/2023-11-…

@marktenenholtz True. But it does a decent job of formulating a LP and MDPs.

One thing I've found ChatGPT is not good at: Solving optimization problems. I swear, I've never gotten anything useful from the optimization code that it writes.

@Aki__Singh @raydistributed @databricks Changes in API: xgboost_ray vs XgboostTrainer. Docs is not easy to navigate. I also find inter operability between ray dataset and spark dataframes buggy. Last, community is slow to respond to questions.

@vbabashov @raydistributed @databricks what challenges you faced?

Anyone using @databricks and @raydistributed in production ?

@hey_madni GPT

Prompt engineers make $200k+ yearly. That's why I built "GPT-4 Prompt Engineering Guide": • Zero-Hero Full Guide • 100+ Examples • Tips, Tricks, Techniques & more. And for 24 hrs, it's 100% FREE! To get it, just: 1. Like 2. Reply "GPT" 3. Follow me (so that I can DM you)

@marktenenholtz CV is not a viable option always. Big dataset -> expensive training on folds.

3. It's really hard to not overfit Actually it's pretty easy if you set up proper CV. Problem is that people use early stopping as a crutch (which causes overfitting). You wouldn't use a different learning rate per val split, why would you change your boosting rounds as well?

Misconceptions about XGBoost: 1. It's hard to tune the hyperparameters Disagree, there's only ~5-6 to tune, and 3-5 of them can be tuned very quickly. Learning rate is probably the one that requires the most care. 2. Inability to extrapolate is a problem (cont)

@marktenenholtz Which 5-6? Can you elaborate? To me it’s actually 8-9

@marktenenholtz I didn’t get an improvement in run times for Xgboost on GPU. Not sure why?

Here’s my general recommendation for using LightGBM vs. XGBoost vs. CatBoost. Use: • XGBoost when you have a GPU • LightGBM when you’re only using a CPU • CatBoost with a lot of categorical features There’s a lot of wiggle room in there, though. The best part: they all thrive with the same general kind of features, so it’s really easy to swap them in and out. For time-series forecasting, start with rolling features, lag features, and go from there.

Operations Research Ranked #1 Highest Earning Field 📢 🚨 College majors have a big impact on income. Here are the highest- and lowest-earning fields [CBS News] cbsn.ws/3MBRGUj #orms #operationsresearch #managementscience #datascience #ML #AI #optimization #informs

Hi @rasbt: What is the right approach to define distribution for the search space of max_depth parameter in xgboost or other tree-based algos. hp.choice() or hp.quniform distribution? I have seen people you use both in Hyperopt trials, just wanted to ask your take on this?