Nice Video brother. As Someone who has been doing WebD + DSA (for fun)+ML. My daily schedule itself is very taxing.At first ,I thought im overloading my brain with too much knowledge .But after being in this journey for more than 6 month and taking every thing slow and consistent ,everything is getting sorted out.
@@PhongGT dude where can I contact you in personal I mean am willing to put in the work daily ,I need a little push or guide ,I can do as you say, i hardly need this am in undergrad too.
I have a general question on the topic. lets say I want to implement a ML model at my work. lets say its predictivive analysis just like this tutorial where given a dataset want to predict at some level of confidence a hypothesis. i would like to know if you can do a video or if its simple enough respond here on the architecture needed to make this happen. is it money will have to be spent on licenses, large scale architecture etc. I need to know before i pitch it to the bosses and continue on my learning journey. thanks in advance
You'll need to start with the data: - Do you have the data? - If you do, how much data do you have? The more data, the more potential to be accurate, but also more expensive to train since it will need to go through more data - If you don't, how will you get the data? That's when you can answer your licensing and $$$ questions - Is it a simple target you're trying to predict? If so you can easily train on a local machine and save the model, if not you could use Google Vertex AI to train, or some other cloud provider that offers a GPU (or just use a CPU if it's a simple enough model). The way I pitched to my bosses was to first get a quick understanding of ML through these competitions, and then trained a local model with existing project data as a POC, then demo'd it to my bosses to say "we can improve this one KPI like this, here's the preliminary results that we can keep pushing even further if we put effort into this". You don't need to know everything that's happening under the hood, but you should probably come prepared with how ML can make an impact vs. the efforts involved.
@@PhongGTi have access to the transactional data. i can take that and clean it etc. does a model take up lot of storage to be done locally? if they are not comfortable with putting data on cloud what will be on prem options? thats what im trying to figure out
@@swankyshivy for transactional data, if you can extract some sort of concrete couple of structured columns and a target column, you can probably train locally or through google colab just drop a csv file after you've figured out how you want your dataset to be, then train with it. models aren't going to be too big at this stage, so you can save it locally as well and save to to whatever on prem machine you got or you can build an on prem training machine, you really only need a modern CPU and a training framework (tf or pytorch) and save it on the same machine try out one of the kaggle competitions and you'll kind of get an idea the size of everything and all the basic components to train
![How does machine learning work? (for beginners) [2024] | ML Starter Pack](http://i.ytimg.com/vi/FqK3ZZlqwfM/mqdefault.jpg)
Nice Video brother. As Someone who has been doing WebD + DSA (for fun)+ML. My daily schedule itself is very taxing.At first ,I thought im overloading my brain with too much knowledge .But after being in this journey for more than 6 month and taking every thing slow and consistent ,everything is getting sorted out.
that's great to hear man, it's been a wild ride trying to self learn and I've been seeing more and more people like me
You deserve a sub bro
thanks man, that means a lot
It's the right time for me to learn from this guy ,do post daily bruh going to learn from you ??
I'll do daily shorts but the long vids probably once a week
@@PhongGT dude where can I contact you in personal I mean am willing to put in the work daily ,I need a little push or guide ,I can do as you say, i hardly need this am in undergrad too.
I have a general question on the topic. lets say I want to implement a ML model at my work. lets say its predictivive analysis just like this tutorial where given a dataset want to predict at some level of confidence a hypothesis. i would like to know if you can do a video or if its simple enough respond here on the architecture needed to make this happen. is it money will have to be spent on licenses, large scale architecture etc. I need to know before i pitch it to the bosses and continue on my learning journey. thanks in advance
You'll need to start with the data:
- Do you have the data?
- If you do, how much data do you have? The more data, the more potential to be accurate, but also more expensive to train since it will need to go through more data
- If you don't, how will you get the data? That's when you can answer your licensing and $$$ questions
- Is it a simple target you're trying to predict? If so you can easily train on a local machine and save the model, if not you could use Google Vertex AI to train, or some other cloud provider that offers a GPU (or just use a CPU if it's a simple enough model).
The way I pitched to my bosses was to first get a quick understanding of ML through these competitions, and then trained a local model with existing project data as a POC, then demo'd it to my bosses to say "we can improve this one KPI like this, here's the preliminary results that we can keep pushing even further if we put effort into this".
You don't need to know everything that's happening under the hood, but you should probably come prepared with how ML can make an impact vs. the efforts involved.
@@PhongGTi have access to the transactional data. i can take that and clean it etc. does a model take up lot of storage to be done locally? if they are not comfortable with putting data on cloud what will be on prem options? thats what im trying to figure out
@@swankyshivy for transactional data, if you can extract some sort of concrete couple of structured columns and a target column, you can probably train locally or through google colab
just drop a csv file after you've figured out how you want your dataset to be, then train with it.
models aren't going to be too big at this stage, so you can save it locally as well and save to to whatever on prem machine you got
or you can build an on prem training machine, you really only need a modern CPU and a training framework (tf or pytorch) and save it on the same machine
try out one of the kaggle competitions and you'll kind of get an idea the size of everything and all the basic components to train
@@PhongGToki great. ill do that and im also following along your playlist. patiently waiting on the next video lol thanks a mil
@@swankyshivy np let me know if you have any further questions or you want me to cover something in more detail in a vid
thankyou will try this I was trying to learn ml
I hope to share as much as I can
Am I doing right to learn both Web Dev and ML together??
learning both is great but I suggest you learn one to a good proficiency before learning the other
@@PhongGT thanks for your advice ☺️