Sumanth Manduru

DS Intern @Microsoft || Ph.D'ing @GeorgeMasonU || Ex - Data Scientist @jio || Research Interests : SLMs, NLP ||

[Sumanth Manduru @SumanthManduru]

@ZiyuYao @AmazonScience Congratulations Professor Yao

[Raj Dabre @prajdabre]

Rule #1 of interacting with a PhD student: Never ask them how their PhD is going.

EVER!

[Sumanth Manduru @SumanthManduru]

@niloofar_mire @icmlconf I'd love to inherit that trait! I'm aiming for more ICML papers too! 🤞 #papers #ICML

[Sumanth Manduru @SumanthManduru]

@sumanthd17 May be it's not his original account.

[Alisa Liu @alisawuffles]

I'm joining OpenAI next week!🥹 The job search turned out to be really challenging but also super rewarding, so I wrote a small blog to share what I learned along the way and hopefully make the process a little less mysterious for the next person. https://t.co/6FigSBdenD

[Phosphen @phosphenq]

This 2 hour lecture by Yann LeCun (Turing Award winner) will teach you why the next trillion dollar AI company won't be built on LLMs.

He trashes the $100 Billion LLM race, attacks Musk and Amodei, declares scaling dead.

Bookmark & watch tonight after work, skip to 7:00.

[Sumanth Manduru @SumanthManduru]

@sumanthd17 @NirantK Lol

[Sumanth Manduru @SumanthManduru]

@haha_girrrl A PhD, how many more years ?

[AK @_akhaliq]

Google presents PaperBanana

Automating Academic Illustration for AI Scientists

[Yanan Sui @YananSui]

@iclr_conf ICLR_2026_Workshops_names_20251201

[Paras Chopra @paraschopra]

What I shared with research interns at @lossfunk on how to go about their research projects.

[Sumanth Manduru @SumanthManduru]

Started Data Science Intern @Microsoft working on Windows Developer Retention.

[Sumanth Manduru @SumanthManduru]

Got rejected by @ericsson for the summer internship 2023.
However, thanks for the interview opportunity @ericsson and @EricssonLabs and their team included in the whole process.

This is N th rejection in the process. I will keep trying........

[Raj Dabre @prajdabre]

I regularly get messages like this and while I don't think I'm the best person to answer how to switch to ML, here is what I would do given what I know.

Simple answer: A good balance of theory and hands on is the best solution. Mess up the balance and you will end up with mediocrity.

Longer answer:

1. You need strong fundamentals so you have to atleast have your basics clear on: probability, calculus, linear algebra, optimization (linear and possibly convex), and information theory.
2. Before getting into neural networks, it's good to know the past. Learn about logistic regression, linear regression, SVMs, decision trees, graphical models, etc. No need to go into too much depth but knowing your history helps.
3. Once you are comfortable, jump into basics like word2vec, recurrent neural networks, attention, encoder-decoder models, transformers, etc.
4. Study your loss functions, optimizers, regularizers. Know what hyper parameter tuning is.
5. At this stage you will have a good enough understanding of theory to know what interests you and you will be able to take it from there yourself.

The aforementioned points are mostly on theory but practical knowledge is also crucial:

1. For every topic you study, try to find a GitHub repo or a notebook on @huggingface and see if you can run it. If you are resource constrained then @UnslothAI provide some awesome notebooks and code that can help you run 7B parameter models and more on a colab instance.
2. If possible, don't just run stuff but dig into these notebooks and see what's happening. Without this step you will not be able to fix issues.
3. Reimplement things. Back in 2021, I developed my own training pipeline for sequence to sequence learning (YANMTT). I modded a lot of code and implemented a number of new functionalities. Hack away. Hands on learning is good.
4. Fail a lot. A lot of ML is uncharted territory so be prepared to get lots of negative outcomes.
5. Sometimes things will work and you won't know why. Sometimes they will not work and you won't know why. Some detective skills will help.

Additional tips:
1. Read research papers after you have your basics clear. Try to implement them. For example, learning about the transformers from a blog is nice but also read the paper.
2. Give yourself a break from time to time. This area is too competitive. Frustration is bound to happen. Take it easy.
3. Try to think of real life applications when learning ML. Helps to narrow down into a niche.
4. Have fun.

This is just my perspective. There may be better ways. Hope it helps.

[Coursera @coursera]

Want a chance to get a free backpack from Coursera? 🎒

Simply repost and DM us your full name, address, email and phone number. 😊

(Exclusive to learners in the US, UK and India)

[Sumanth Manduru @SumanthManduru]

@willcb Interested

[Sumanth Manduru @SumanthManduru]

@chrissyykat Hey @chrissyykat I am currently pursuing PhD. I would be delighted to work in Google Deepmind. Please do help me in applying to that role.

[Sumanth Manduru @SumanthManduru]

@adityakusupati @uwcse @GoogleDeepMind I am currently pursuing my Ph.D. (First Year). It would be helpful if you can share your journey right from PhD to Google Deep Mind.

[Sumanth Manduru @SumanthManduru]

@ManeeshKSingh5 I am interested in a job for summer 2024. Could you please dm me as I couldn't do ?

[Sumanth Manduru @SumanthManduru]

@jaeh0ng_yoon @UNC @mohitban47 @uncnlp @unccs Congratulations. Working with @mohitban47 is always on my wish list.

[Sumanth Manduru @SumanthManduru]

@rajammanabrolu @UCSanDiego @ucsd_cse @MosaicML Congrats Prof Raj