논문 PDF만 쌓이고, 읽고 나서도 남는 게 없다면 이 3개 저장해두면 됨.
관련 논문 찾기 → 막힌 내용 질문하기 → 근거와 메모 남기기.
관심 분야를 처음 공부하거나 연구 주제를 탐색할 때 쓰는 조합임.
1. ResearchRabbit
논문 하나를 기준으로 비슷한 연구와 앞뒤로 이어지는 연구를 찾아보기.
“이 논문 다음에는 뭘 읽어야 하지?” 싶을 때 쓰기 좋음.
기본 검색과 논문 모음 기능은 무료임.
researchrabbit.ai
2. Gemini Notebook — 기존 NotebookLM
논문 PDF를 넣고 연구 문제·방법·결과를 질문하기.
이해가 안 되는 부분을 좁혀 묻고, 답변에 붙은 출처로 원문을 확인할 때 쓰면 됨.
notebook.google
3. Zotero
논문과 메모를 함께 보관하고, 밑줄 친 부분을 출처가 연결된 노트로 모으기.
나중에 “이 설명이 어느 논문에 있었지?”를 다시 찾을 때 쓰기 좋음.
zotero.org
논문 찾기 → 질문하며 읽기 → 원문 근거와 메모 저장하기.
처음에는 논문 한 편에서 ‘무슨 문제를 풀었는지, 어떻게 풀었는지, 어디까지 확인했는지’만 남겨보면 됨.
AI 요약을 모으는 것보다, 내가 설명할 수 있는 내용과 다시 확인할 근거를 함께 남기는 데 쓰면 좋음.
Catscan: Visualizing Pipelines of CPU Performance Simulation
Aaron Lindsay, Nicholas Kelly, Scott Witscher, Mahesh Madhav
arxiv.org/abs/2610.02121 [𝚌𝚜.𝙰𝚁 𝚌𝚜.𝙷𝙲 𝚌𝚜.𝙿𝙵]
I put together a handbook on how Mixture of Experts actually works, starting with what happens when a Transformer FFN becomes a routed bank of experts and only a small subset is selected for each token.
It works through router logits, Top-k selection, expert weighting, load balancing, capacity and token dropping, router z-loss, dropless execution, total vs. active parameters, expert parallelism, and how these ideas show up in Mixtral, DeepSeek-V3, and Qwen3. Worked examples throughout, with citations back to the original papers and current implementations.
Made it to be the explanation I wish I'd had when I first tried to understand how modern MoE models can grow so large without activating the full expert bank for every token.
Inference roadmap I gotta follow:
1) Watch this course fully : youtube.com/playlist?list=…
(Covers everything related to LLM Inference+ Training but on high level)
2) Gonna read this resource: jax-ml.github.io/scaling-book/ ( for deep dive on topics)
3) Read PMPP(CUDA & GPU specific chapters)
4) Blogs and codebase of vLLM , SGlang, llama.cpp
5) Inference Engg by philip book for brush up on the concepts
6) Also these notes which i found randomly on X for brushup: x.com/gauri__gupta/s…
Along with it : contributions to the above mentioned repo for understanding depth + talking with folks building in inference here
inference is future.
tried going as much in detail as possible for the writing GEMV kernel in CUDA,
hopefully it will be a good read. ciao!
heyyanshuman.com/posts/making_g…
Excited to co-instruct MIT 6.5940: TinyML and Efficient AI Computing with @songhan_mit this fall!
Looking forward to teaching, learning, and building with an incredible class.
Course site: efficientml.ai
i trained a computer vision model to train me at tennis
it tracks ball speed, forehand vs backhand strokes, shot placement, body position when i strike the ball
how this works:
- input video from an iphone
- CV pipeline orchestrated with roboflow agent: data labelling using Astra 6, detection done with a fine-tuned RF-DETR model
- video annotation using python + openCV
goal is to make an app that records each practice session, tracks progress over time, and gives feedback on improving my technique
more to come!
84 Followers 1K FollowingFounder, CEO & CPO of VENDIT Inc.
We take care of every aspect of hotel property operations and sales — from A to Z.
Raised $7 million in our Pre-A round.
791 Followers 2 FollowingNew Hardware Architecture papers from https://t.co/6hoXZ9RzY2: systems organization and hardware architecture. Thank you to arXiv for use of its open access interoperability.
342K Followers 66 FollowingWe're sharing/showcasing best of @github projects/repos. Follow to stay in loop. Promoting Open-Source Contributions. UNOFFICIAL, but followed by github
2K Followers 3K FollowingResearch Scientist@#GoogleDeepMind AIxCC/DEF CON CTF Champion. My tweets do not reflect the view of my employer. A sponsor of 🐻
1K Followers 7K FollowingSomewhere between machines and people. Less is exponentially more. Deciding what not to do is as important as deciding what to do. 靑天亂流.
222K Followers 3K FollowingFollow for posts about GitHub repos, DSPy, and agents
Subscribe for top posts
DM to share your AI project (Due to volume of DMs I'll prioritize subscribers)
607 Followers 603 Following이중언어 웹페이지 번역 사이트
웹·PDF·ePub·자막 무료 번역 | 최상급 품질
ISO 27001·27701 국제 인증
@ProductHunt
#buildinpublic #indiehacker
@immersivetransJ @immersivetrans