Suresh (@_Suresh2) header
Suresh (@_Suresh2) profile picture

Researching AI for Software Engineering, LLM-assisted program analysis, API compatibility, framework evolution | MS @ Chongqing University

Chongqing China
Joined January 2019
Photo by Suresh (@_Suresh2)Photo by Suresh (@_Suresh2)Photo by Suresh (@_Suresh2)Photo by Suresh (@_Suresh2)Photo by Suresh (@_Suresh2)Photo by Suresh (@_Suresh2)Photo by Suresh (@_Suresh2)Photo by Suresh (@_Suresh2)Photo by Suresh (@_Suresh2)Photo by Suresh (@_Suresh2)
Suresh retweeted
搞科研的都知道,画图比跑代码还折磨人。 数据跑完了,卡在配色、字体、排版和架构图上,能让人把头发全薅光。 今天挖到四个搞科研绘图的神器,建议直接放进收藏夹吃灰。 第一个,Scientific Visualization。 这玩意儿简直是期刊质检员附体。以前画个柱状图,总因为误差线没标 SEM 还是 SD 被审稿人怼。 它直接内置了顶刊规范,色盲友好和黑白打印模式全都兼容。 最关键的是,导出的 PDF、EPS 和 TIFF 直接满足分辨率要求,省去了反复返工的痛苦。 github.com/Panda-dda/Scienti… 第二个,Academic Research Skills 里的 Visualization Agent。 数据跑出来了不知道配什么图最合适,它能根据数据特点选图,并直接生成能运行的 Python 或 R 代码。 连 LaTeX 插图代码都一并给你写好,绘图和排版一步到位。 github.com/Imbad0202/academi… 第三个,Nature Figure。 专门治几张子图拼在一起像大杂烩的毛病。它会先问你核心结论是什么,再让每张子图各司其职。 生成后自动检查文字遮挡和边距对齐,对不齐直接不让你导出。 主打一个证据链清晰,让外行一眼看懂。 github.com/Yuan1z0825/nature… 第四个,深度之眼开源的 CCF-Figure。 AI 论文的架构图最让人头疼。它会先判断你的论文属于七类中的哪一种,再从十一种结构里匹配布局。 架构闭环和反馈分类都能安排,最后输出中英双语提示词,直接拿去给大模型出图,还附带了翻车自检清单。 github.com/Deepshare-Officia…
7
171
2
1,135
57,336
英語論文執筆におけるAIっぽい表現を修正するスキル”Humanizer academic”マイナーアップデート github.com/matsuikentaro1/hu… 最近のGPTやClaudeはcoding agentとして最適化された結果?人間には読みにくい言葉遣いが多い…このスキル使うとちょっとマシになります(中華圏でよく使ってもらえてるっぽい)
7
82
1
770
55,029
ETH Zurich(苏黎世联邦理工学院) 把 2026 年整门机器人学习课公开了。不是精简版网课。 是正课本身:课件、课堂录像、编程作业、GitHub 仓库全在。 课程从模仿学习、强化学习一路讲到 Vision-Language-Action 模型和机器人基础模型。 客座包括 Physical Intelligence 联合创始人 Quan Vuong、Diffusion Policy 作者 Cheng Chi、Dieter Fox 等一线研究者。 12 周。 免费。 不用注册。 主讲人是 Oier Mees 和 ETH Zurich 团队。想搞清楚机器人智能真正在往哪走,这就是这个领域现在在用的那份清单。 课程首页:cvg.ethz.ch/lectures/Robot-L…
2
66
1
287
17,452
Hi all, I wrote a guide to applying to PhD programs based on my own experiences. I hope it is a helpful resource! pz1729.substack.com/p/guide-…
12
104
1
857
56,094
Suresh retweeted
We just killed Exa, Tavily, SerpAPI, and Brave. Your agent can now search & fetch any webpage for 100% FREE. Them: $7 per 1,000 searches. Us: $0. No subscriptions, no quotas. Humans search Google for free. Agents shouldn't have to pay either. Made possible by @Tiny_Fish and @MonidHQ.
387
349
134
4,746
892,491
Suresh retweeted
50 websites that feel like the internet’s hidden toolbox 🧰 1. unpaywall.org — Free research papers 2. openlibrary.org — Borrow books online 3. doaj.org — Free academic journals 4. alternativeto.net — App alternatives 5. justwatch.com — Find where to stream 6. archive.org — Internet archives 7. gutenberg.org — 70K+ free books 8. openstax.org — Free textbooks 9. openculture.com — Free courses 10. wolframalpha.com — Solve complex problems 11. photopea.com — Photoshop alternative 12. squoosh.app — Compress images 13. remove.bg — Remove backgrounds 14. cleanup.pictures — Remove objects 15. unscreen.com — Remove video backgrounds 16. carbon.now.sh — Beautiful code images 17. ray.so — Code screenshots 18. shots.so — Product mockups 19. smartmockups.com — Create mockups 20. haveibeenpwned.com — Check data breaches 21. virustotal.com — Scan files & URLs 22. privnote.com — Self-destructing notes 23. temp-mail.org — Temporary email 24. file.io — Temporary file sharing 25. archive.ph — Save webpages 26. similarsites.com — Find similar websites 27. radio.garden — Explore global radio 28. everynoise.com — Discover music genres 29. tunefind.com — Find songs from shows 30. musicforprogramming.net — Focus music 31. mynoise.net — Custom background sounds 32. coffitivity.com — Café ambience 33. elicit.org — Research assistant 34. consensus.app — Research-backed answers 35. connectedpapers.com — Research connections 36. semanticscholar.org — Academic search 37. scispace.com — Understand research papers 38. summarize.tech — YouTube summaries 39. phind.com — AI for developers 40. regex101.com — Test regex 41. codebeautify.org — Format code 42. jsonformatter.org — Format JSON 43. explainshell.com — Understand terminal commands 44. raindrop.io — Bookmark manager 45. downdetector.com — Check outages 46. tineye.com — Reverse image search 47. fast.com — Internet speed test 48. smallpdf.com — PDF tools 49. ilovepdf.com — Merge/split PDFs 50. 10minutemail.com — Temporary email Save this. You’ll definitely need some of these later. 🔖 Follow @I_AM_AYASHA for more useful websites, AI tools & tech resources. 🚀
20
399
5
1,320
59,691
If I were hiring a PhD student today, these are the emails I’d reply to. Most emails begin by telling me how much they admire my research. That doesn’t help me. Tell me which assumption in my recent work caught your attention. Tell me which limitation you’d want to test. Tell me where your own research unexpectedly overlapped with mine. Within a few sentences, I know whether you’ve read a paper or understood a research programme. The second thing I look for is whether you understand your own work. Don’t tell me you used CRISPR, RNA-seq, or machine learning. Tell me why you chose that method over another. Tell me what failed. Tell me what you would do differently if you repeated the study. Methods can be taught. Scientific judgement cannot. Finally, I ask myself one question: “Can I already imagine this person discussing data at our lab meeting?” Very few emails answer that question. The strongest ones don’t sound like requests for supervision. They sound like the beginning of a collaboration. That’s why they get a reply. If your emails aren’t getting responses, the problem usually isn’t your English or your CV. It’s that you’re introducing yourself instead of introducing your thinking. DM me “Email” if you want me to review your outreach the way a PI actually reads it.
22
99
11,224
Claude Code alone is just a smart intern. These 13 GitHub repos turn it into a full product team. Every repo is public and free. Memory, research, browser automation, QA, and design. Core system Superpowers → github.com/obra/superpowers Context7 → github.com/upstash/context7 Claude-Mem → github.com/thedotmack/claude… Anthropic Skills → github.com/anthropics/skills Loops Ralph Wiggum → github.com/anthropics/claude… Research Crawl4AI → github.com/unclecode/crawl4A… MarkItDown → github.com/microsoft/markitd… Browser + QA Playwright MCP → github.com/microsoft/playwri… Agent Browser → github.com/vercel-labs/agent… Browser Use → github.com/browser-use/brows… Design shadcn/ui → github.com/shadcn-ui/ui Magic UI → github.com/magicuidesign/mag… Motion Primitives → github.com/ibelick/motion-pr… You don't need to install all 13. Pick the one bottleneck slowing your agent down. Bookmark this before you build your next project.
4
6
37
4,783
I drafted a list of RL environment startups. Calling it Pavlov's List (1/2): AfterQuery. Code & Finance. @CarlosGeorgescu, @spencermateega, Danny Tang BenchFlow. Code. @xdotli Bespoke Labs. Enterprise. @madiator, @AlexGDimakis Calaveras. Code. @cis_female, Alana Xiang Cua. Code & Computer Use. @francedot, @ddupont808 Collinear. Enterprise & Code. @nazneenrajani, @soumyadeepb_ d_model. ML & Alignment. @dlbydq, @dmooooon Datacurve. Code. @serenaa_ge, @charleyslee Deeptune. Enterprise. @timlupo Fleet AI. Enterprise. @nicoup General Reasoning. Long Horizon. @rosstaylor90, @ChengxiTaylor, Kip Parker, Thomas Grady Halluminate. Long Horizon & Finance. @Jerr_Wu, @wgm752 Habitat. Code & Computer Use. @maxim_enis, @maxkan, @AndrewMegalaa Haladir. Code & Math. @jibranhutch, @quanmhuynh, @preston281s, @josephtso914 Hillclimb. Math. @jparkjmc, @agithief
10
25
8
313
50,618
> be you > want to actually learn how LLMs work > sick of “just start with linear algebra and come back in 5 years” > decide to build my own roadmap > no fluff. no detours. no 200-hour generic ML playlists > just the stuff that actually gets you from “what’s a token?” to “I trained a mini-GPT with LoRA adapters and FlashAttention” > goal: build, fine-tune, and ship LLMs > not vibe with them. not "learn the theory" forever > build them > you will: > > build an autograd engine from scratch > > write a mini-GPT from scratch > > implement LoRA and fine-tune a model on real data > > hate CUDA at least once > > cry > > keep going > 5 phases > if you already know something? skip > if you're lost? rewatch > if you’re stuck? use DeepResearch > this is a roadmap, not a leash > by the end: you either built the thing or you didn’t > phase 0: foundations > > if matrix multiplication is scary, you’re not ready yet > > watch 3Blue1Brown’s linear algebra series > > MIT 18.06 with Strang, yes, he’s still the GOAT > > code Micrograd from scratch (Karpathy) > > train a mini-MLP on MNIST > > no frameworks, no shortcuts, no mercy > phase 1: transformers > > the name is scary > > it’s just stacked matrix multiplies and attention blocks > > Jay Alammar + 3Blue1Brown for the “aha” > > Stanford CS224N for the theory > > read "Attention Is All You Need" only AFTER building mental models > > Karpathy's "Let's Build GPT" will break your brain in a good way > > project: build a decoder-only GPT from scratch > > bonus: swap tokenizers, try BPE/SentencePiece > phase 2: scaling > > LLMs got good by scaling, not magic > > Kaplan paper -> Chinchilla paper > > learn Data, Tensor, Pipeline parallelism > > spin up multi-GPU jobs using HuggingFace Accelerate > > run into VRAM issues > > fix them > > welcome to real training hell > phase 3: alignment & fine-tuning > > RLHF: OpenAI blog -> Ouyang paper > > SFT -> reward model -> PPO (don’t get lost here) > > Anthropic's Constitutional AI = smart constraints > > LoRA/QLoRA: read, implement, inject into HuggingFace models > > fine-tune on real data > > project: fine-tune gpt2 or distilbert with your own adapters > > not toy examples. real use cases or bust > phase 4: production > this is the part people skip to, but you earned it > inference optimization: FlashAttention, quantization, sub-second latency > read the paper, test with quantized models > resources: > math/coding: > > 3Blue1Brown, MIT 18.06, Goodfellow’s book > PyTorch: > > Karpathy, Zero to Mastery > > transformers: > > Alammar, Karpathy, CS224N, Vaswani et al > > scaling: > > Kaplan, Chinchilla, HuggingFace Accelerate > > alignment: > > OpenAI, Anthropic, LoRA, QLoRA > > inference: > > FlashAttention > the endgame: > > understand how these models actually work > > see through hype > > ignore LinkedIn noise > > build tooling > > train real stuff > > ship your own stack > > look at a paper and think “yeah I get it” > > build your own AI assistant, infra, whatever > make it all the way through? > ship something real? > DM me. > I wanna see what you built. > happy hacking.
20
73
1
1,263
90,390
the thought that DeepMind and UCL have perfect 10 lecture playlist on youtube and i don’t have time to watch it makes me feel guilty as hell.
9
87
1
1,022
41,508
Stanford's Courses on AI & ML (FREE): ❯ CS221 - AI ❯ CS229 - ML ❯ CS229M - ML Theory ❯ CS230 - DL ❯ CS234 - RL ❯ CS236 - Deep Generative Models ❯ CS336 - LLM from Scratch ❯ CS224N - NLP with DL Course links inside:
21
234
5
1,770
98,365
🎉 Happy Labor Day from Genspark! Win 1 Month FREE Genspark Plus membership! HOW TO ENTER : - Repost this post - Follow @Genspark Ends Sept 1, 11:59 PM PST. 10 WINNERS announced Sept 2. Take the day off, Genspark’s on it!
42
178
21
169
26,813
Introducing Devv The first coding agent that helps you ship AI-powered products. Think n8n + Lovable, but natively integrated with Auth, Database, Email, LLMs, OpenRouter, Replicate, Stripe and more. Repost + Reply to get 1-week free credits.
303
302
60
1,006
280,295
holy sh*t. these 10 GitHub repos are a crash course in AI agents & MCPs. you'll go from “what’s a tool call?” to building fully autonomous systems. these are gold mines:
15
68
2
816
112,339
18 useful GitHub repositories every developer should bookmark (you will be thankful):
196
967
16
3,377
588,749
YouTube Courses to Learn and Master HTML and CSS A Thread🧵
32
341
4
1,185