Not sure what ML paper to work on? Feel free to grab my FREE cheat sheet on different ML domains and open challenges: borismeinardus.substack.com/p/a-list-of-different-ml-domains
Hey man your videos are awesome, can you do a part 2 for this video explaining how to start reading papers as a beginner or if possible give an order of topics ranking them from easy to implement to the hard ones which might need more experience. If you can do that it would be really great.
Great suggestion! I actually already have a video on how to read a paper :) ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-e8daXxP-28c.html Regarding the list, it's difficult to do that because there simply are waaaaay too many papers haha But I'll keep the idea in mind 😊 You can here download a free cheatsheet I made to get an overview of the different ML domains boris-meinardus.ck.page/2f5e05bb1f If you find something interesting, you then just need to google that topic. Perhaps it is a good heuristic to say that older papers are a bit simpler than more recent ones. I hope this somewhat helps! 😊
Actually I am a 2nd year student ug cse and I am currently doing DSA I literally don't have any interest but still for the sake of doing I am doing I had intrest in full stack web development but as the saturation increased I personally felt this is not for me I even had tried a bit so now my main goal is Machine learning (AI/DS) I have 2 questions for you mainly 1. Is it very hard to crack ml jobs or any gen ai jobs as a btech student like they often say you need PhD masters etc etc so how hard it is and how possible is it 2. I got a basic idea about we have to learn maths , python , ml,dl, etc etc But still as you are a engineer in itself which is a very great opportunity for me to learn from such a mentor I want an crystal clear explanation from where to start ,how to start and how would you start if you were me ? I am really greatful and thank you if you have read this message . Please feel free to reply in your own way. Thank you.
Hey! So, first of all, I am an ML researcher and not really an engineer. Although they do have an overlap :) The level of education varies a lot based on the company and role. For researchers, you will commonly need a PhD or a masters, but there still are ways to get a researcher job without that. It will just require the equivalence of experience, e.g. at a job. Depending on your skill and display of that (through projects you have worked on) you can even get Data Science or ML engineering roles without a degree. But that might mean to first start at a smaller company and then build expertise to switch to a more prestigious company. The companies look for certain skills and proof for those skills. How you get them (college, self taught or boot camps) is of less and less importance. But again, for research specifically, you need some sort of research experience. That can also happen by starting as a ML engineer and working as an engineer on a research project and then internally pivoting to a research position. There are many different ways to get into ML and Data Science. I hope this somewhat helps :)
@@borismeinardus thank you so much I have been following your content recently from few weeks it's great keep continuing and share some good resource for the machine LEARNING and data science related stuff thanks a lot from student community
Great video! Came here from another one, where you talk about „recreating a paper“, because I didn’t understand what that meant 😂 Was really surprised to find an answer right in the next video I open.
Perfect timing Boris! I was just reimplementing Show, Attend and Tell and I was just starting to lose my mind😅 This video gave me the push I wanted. Btw, just out of curiosity, which paper did you reimplement that got you an interview at DeepMind?
hi, my 2 years of masters will complete in 3 months i have been specializing in AI and Cyber Security. in 1.5 years i have explored software development extensively and now i have to make a career in ML as a Researcher or Engineer My Skills: Data Science and AI => Begineer Python => Intermediate Software Development => Begineer to Intermediate can you guide me how to get start in ML.
Hmm, difficult to give a detailed answer here in the comments, but for short, try to work on cool projects and try to target ones that might be interesting to very specific company projects or even better people. If you can‘t do that in your spare time, perhaps it might be worth considering a PhD. But yeah, try to really think in terms of being very good at one specific things a specific team or person might be looking for. I hope this helps :)
Great video! Would love to see you do this end to end either on a YT Live or a pre-recorded version, this would greatly help people who have never implemented it. Can you also share a sample github of your implementation that got you the deepmind internship?
Thanks! This is a very extensive project, it‘s really tricky to put into one video or one live stream 😅 here is my repository: github.com/sudo-Boris/stable-baselines3 And I sadly didn‘t actually get the internship - just got the interview and then sadly didn‘t pass 😬
I hope one day you will post a video / technical blog on re implementation of research papers. Thank you so much for the GitHub, will surely learn a lot from it
I have even built an new architecture for my research paper. I'm a final year CSE student from India. Still I'm not getting any interview calls. 😢 Can you give me some tips on how can I showcase my skills to the recruiters.
At some point, when you feel confident in you skills and have 1 or 2 "significant" projects, it's all about sales. You need to learn how to sell yourself properly. I have a video on how to write a good resume. That might be a very good point to start! :)
can you suggest me websites of research paper and how you find the research paper of any projects?...btw i like ur videos and the way u explain all major and minor things , thank you!
Thank you! I'm always trying out different things to improve :) Well, technically almost all papers are on arxiv.org but I would simply suggest to google an area you are interested in and append the word "papers".
Find something that interests you. There are many different domains to explore (in the pinned comment you can get my free little cheatsheet on different domains). Once you found an interesting domain or problem, simply start googling that concept and add „papers“ :) I think thinks a very good way to start.
Really nice video Boris! Do you think re-implementing from scratch some basic models, either like CNN or GPT models can be useful both for stand out and understanding?
Hey Boris, Incredible video! Just wondering to work at DeepMind do you have to have a PhD. I see so many jobs in ML and AI requiring PhD's in related fields these days
Thanks! Yeah that‘s a tricky question. The short answer is, you theoretically don‘t, but you somehow need equivalent experience. I will elaborate on this a bit more in this weeks newsletter ☺️ but the main point is, many AI/ ML positions require very domain specific knowledge, and you usually get that by studying a very specific problem and working on it for years. This is something you do in a PhD. I hope this already helps a bit :)
@@borismeinardus Ah thank you so much for your reply. So when you were applying for the internships did you have a PhD at that point? Also, in regards to reproducing papers, do you have any tips for finding ones that aren't too complex?
Hello sir can you please help with one problem Develop a framework for leveraging machine learning to uncover customer trends and behaviors while adhering to the highest standards of ethical data usage. In today's rapidly evolving business landscape, understanding customer preferences is crucial; however, privacy concerns have taken precedence. We are actively seeking innovative solutions to anonymize user identities within extensive databases, enabling the responsible application of machine learning techniques. Your objective is to construct a robust and privacy-centric framework that enables organizations to extract valuable insights without compromising individual rights.
You just describe what you did "reimplement a paper and recreate results of original paper". Perhaps add one or two more key details to have some important key words on your resume. Perhaps have a look at my resume video. I there also have my paper reimplementation project on my resume :)
Can anyone tell me about amazon ML applied scientist intern role...Like what should I specifically focus on and how can I increase my chances for getting selected..I been into ML domains for like 6-7 months and currently studying depth working of VAE's
Are you able to explain why this is a good approach? This seems a bit un-original to me, and if authors also released code then I don't think you could impress many people by doing a re-implementation.
There are multiple points: - It is a challenging project and you demonstrate almost all relevant skills of working on a ML project - It shows the employer that you have some expertise working on a project that can be directly applied to their work (depending on der overlap between your paper and the company you are applying for) - Especially for research, recreating results from other baselines is a fundamental part of your final work (comparing performance of different ideas/ approaches) - even if there already is a public repository with the implementation, if you work on your own (ideally without looking at their code) you still go through the process of understanding and developing. When in an interview, you can then really discuss the process, your experiments, and results. That is what the company is looking for. - This work builds the foundation on which you can theoretically build upon. You can then try to do further experiments and see if newer techniques improve the performance. That is then even more next level and you can say you didn‘t just recreate the results but even improved them I hope this somewhat helps and grounds my reasoning that this project is arguably the best one you can work on (on your own, not taking actual crazy work experience into consideration)
Hmm, that's a tricky one because it is always really different depending on the type of research you are doing. But I will keep that in mind and do my best!
Hi Booris , I have recently, got an internship in field of machine learning San LLM intern . I am making models with help of langchain and hugging face open ai , google , lama 2 . I am 1st year of my graduation . How should I create more project in field of LLM like fine tune with pretrained data . And yes I have recreates project video lama 2 . It was not very good but t helped me to get an internship . Looking for more opportunities in this field .
Sounds great! I honestly think you are doing well and if you continue doing what you are doing you will definitely get somewhere :) But to perhaps give you an a bit more satisfying answer, I think you should try to extend what you have already done. Make the video llama work better, cleaner, and write a little technical report/ blog post about it. You can also look into contributing that work to Hugging Face, or contribute otherwise. I think if you have one more great project, you need to focus on packaging it properly. I hope this somewhat helps :)
Depends on the final role you are applying for. But in general, you do need a good understanding of linear algebra, statistics, and a bit of calculus. In practise, you will likely not need much, but to learn everything, you need to have that understanding and ideally enjoy studying the math part.
haha thank you I guess growing up watching movies and shows in english helped and reading books in english, reading ML papers in english, and speaking a lot of english with colleagues also helps haha ☺️💛
@@kolyakravchenko7086what? Why are you saying that. He is delivering a constructive criticism about the sound quality, which I agree with because the sound is quite low. Furthermore, it helps to develop the quality of the content.