soham R.

🔐 New Paper Out! "Guardians of Generation (GoG)" introduces a model agnostic framework that prevents diffusion models from reproducing copyrighted content, without any retraining! Balances creative freedom with copyright compliance. Check out our work: arxiv.org/abs/2503.16171

Here is my undefeated hiring strategy: - Neurodivergent - Consistently works out/health conscious - Has reached top 1% in any video game - Uses claude Stg this is the high performer blueprint.

Woman in the gym took my 15kg dumbbell that i was using for wrist curls And proceeded to use it as a phone stand for video call

My paper got 47 likes on Alphaxiv. Cool stuff! alphaxiv.org/abs/2503.16171

Cisco SWE Interns for the summer hmu

@thestoicccoder 5 A*s by 22 is insane. Could you share some of your research work perhaps?

my museum of victories as a 22 year old: 1. working remotely as a research scientist in post-training/multimodal ai with a ctc of 78k usd which is approx 72 lpa INR 2. visited 7 countries: singapore, japan, china, hawai, thailand, vietnam, south korea . worked remotely from a few others were company offsite 3. graduated with more than 7 off-campus job offers all over 18lpa from a college where highest package was 10. 4. completed over 11 internships spanning research, sde, ml and devops 5. 5 publications in A* conferences including ICML, ICLR, NeurIPS 6. worked with some of the top professors and phd students in the world getting some insane LORs in the process 7. got ms admits from 7 of the top 15 US colleges 8. Used to weigh 52kgs when i started engineering, got to 79kgs with a decent amount of muscle. 9. helped rebuild my grandparents house. 10. went on a full north to south vietnam trip with my bestfriend of 15 years 11. helped pay off parent's debts 12. finally bought a MacBook of my own 13. working on my first solo author paper under guidance of someone from DeepMind the only reason I posted all this is because I know how scary it is getting overwhelmed by all the progress happening and it's okay to be scared that you might not make it. but there's something you should remember: It only takes one big win to cancel all the losses. just one win. for me that win was getting my first internship and it was just uphill from there. I'll say the same thing I said a year ago, keep your head down, keep working, keep learning , do not lose that curiosity at any cost, have that consistency, it's fine to have off days, it's fine to get distracted but never lose sight of your end goal. and when you do win you'll realise the amount of effort it actually takes to make it and you'll be grateful and proud of yourself that you put in the work. may you never ever lose the desire to pursue things no matter how hard it gets. keep grinding. ps: a photo from my first international flight

Things have been going quite well recently. Feels like the calm before the storm :)

The only way to not become an unemployable cow after 2-3 years from now is to divert ur focus from building(with claude), and back to learning from first principles.

@TukiFromKL Memory exists in the application layer. LLMs do not “remember” your 2 month old conversations. Infact, they don’t “remember” anything unless you keep feeding it into their context. You’re looking at a rather trivial engineering problem here, slop account with 100k followers.

🚨 Are you paying attention to what Karpathy just admitted.. the founding member of OpenAI.. the guy who trained the models you use every day.. just said every single LLM has the same problem.. ask it one question two months ago and it treats it like your entire identity.. oh wait.. you mentioned crypto once in January? congratulations.. you're now a crypto guy forever.. you also asked about a recipe? every conversation starts with "as someone who enjoys cooking.." these models don't remember you.. they stereotype you.. off a single data point.. we gave AI a photographic memory and forgot to give it the ability to forget.. and forgetting is the most human thing there is..

Openreview hacked before ARR deadline

@underDawg03 @Ra1kshit Especially that dristibution woman

@Ra1kshit Yeah i dont like the people of matiks because they are too pretentious but idt there was anything wrong with that post

i dont get the hate here, his stipend is not 5K anymore, and afaik matiks wasn't even funded at that time. this isn't a ragebait either, y'all just need to calm down

@cneuralnetwork But is that a good thing or a bad thing :( Most people wouldn’t want to settle for their PPO

accenture took 44 people from my college for intern 😁

what's the use of working hard and doing good if no one is proud of you and genuinely no one cares about you? remain a failure and be happy

Stop “fixing wording and tone” of your posts with LLMs, especially chatGPT. Ive got a headache from reading posts sounding exactly the same all over my tl.

@_adityaa21 I get it. My point was that when the wording and tone comes from ai, it loses personal authenticity and individuality and makes everyone sound the same.

@neuralxtract Fair point. I use AI to refine wording, not to generate opinions, the thoughts are still mine. Idea is mine, AI just helps structure it faster.

Companies Visiting Tier-1 vs Tier-3 Colleges 🎓 Placement season in India is basically two different Netflix shows depending on your college tier. Tier-1 Colleges Day 0: Google, Microsoft, Amazon, Adobe, Uber, Atlassian, Goldman Sachs enter campus like it’s a royal wedding procession. Students: > “I’m confused… should I take ₹45 LPA or wait for ₹60 LPA?” Companies: > “We only need the top 200 students.” College: > “We have 1,800 of them.” Coding rounds feel like competitive programming finals. HR round feels like a podcast conversation about “impact”. Offer letters come with relocation bonus, joining bonus, stock bonus, happiness bonus, breathing bonus. Rejection reason: > “You were very strong, but someone solved the graph problem in O(1).” But in Tier-3 Colleges - Day 0: “Random IT Services infotech Pvt Ltd” arrives with a logo designed in MS Paint. CTC: ₹4.5 LPA In-hand: ₹18,200/month Bond: 2 years Penalty: ₹2 lakh Students: > “Is it WFH?” Company: > “Yes — Work From Hub in another state.” Coding round: Reverse a string, print star pattern, explain why you love the company you heard about 7 minutes ago. HR round: > “Are you comfortable with night shifts, rotational shifts, weekend shifts, festival shifts and life shifts?” Role offered: Software Engineer Actual work: Update Excel sheets + call clients + sometimes restart the router. Meanwhile Parents- Tier-1 parent: > “Beta don’t focus on money, focus on learning.” Tier-3 parent: > “Beta take anything, learning later.” After 6 months- Tier-1 graduate: debating startup vs higher studies vs US relocation Tier-3 graduate: debating developer vs support vs UPSC vs YouTube vs farming Same syllabus. Same university exams. Same 75% attendance torture. Different universe. India doesn’t have one placement system. It has parallel economies running on the same degree.

4000 goats were herded in a fence for 4 years. Most grazed peacefully and enjoyed the fodder. Some tried to be smart and train hard to jump over the fence for 3 years. some succeeded. Most didn’t. In the end, nothing changes the fact that they were all goats.

It is of grave misfortune that I possess the ability to mog every individual in my batch in every aspect of human endeavour academic/non-academic. Your environment should be one where ur ego gets shattered with every passing second. I hope I get to such a place one day.

@cattoiscoding dead internet theory will be real very soon.

I have been browsing X for the past 20 mins or so and every fucking post I came to was AI generated. Shit is so obvious : "Its not just mediocre, its extraordinary" "One_Word_Question? One_Word_Answer." Fake scenarios and other bullshit.

Midsems in 4 days. Can i repeat this with 40% attendance?

@kabir_j25 @ponguru Congrats

I have joined precog@IIITH as a research intern. Will be working under Prof. @ponguru for the next 6 months :)