Sumit Vishwkarma (@sudo_core) header
Sumit Vishwkarma (@sudo_core) profile picture

Linux and coffee

Joined April 2024
Guess, it would've been incomplete without my cameo ! #chaicode #chaicohort
3
2
33
713
My dad had a problem with managing quotations so I built an app for it. lowkey, a simple little thing.. but makes the work easier.. also shoutout to @ChaiCodeHQ and @surajtwt_ for the mobile dev cohort 🙌🏻
2
1
23
672
At Devfest 2026 Crazy event by @gdg_noida Last time I was here as a newbie, this time as a Developer .. ! Lessgoo #devfest #gdgnoida #gdgcloudnoida
2
1
32
704
I was always wondering.. that okay graphs are made for so many things by LLM, but how does it find the stuff.. SQL query? No SQL? Isn't it slow? Then , I got introduced to GraphDB It was a fun class Notes : inapp.app/sudo/gen-notes Huge shoutout to @piyushgarg_dev #chaicode
1
25
597
I Built AI applications before I truly understood what was happening inside an LLM AI jargons were familiar but not clear to me So I sat down , connected all the dots and here's byproduct of my understanding A video - Me & canvas youtu.be/l7vLKUevapg?si=6oss… #BuildInPublic
1
2
20
723
Did you know? That LLMs can also dream? Yes they can When you're building a memory engine for ur ai application. It's crazy how close the system tries to mimic human brain. Here's my notes from "Memory Engine" class by @piyushgarg_dev PDF : inapp.app/sudo/gen-notes #chaicode
2
44
641
RAG RAG RAG , AI AI AI enough! I also wanna play presenting DOX - connect your documents, research notes, and creative thoughts @piyushgarg_dev sir told me to build this , in our recent class . I said why not check it out : dox.sud-o.app #chaicode @ChaiCodeHQ
3
54
1,497
anyone can write code that works. the fun begins when you start asking why it's slow. understanding system design, batching, parallelism, and bottlenecks turned today's pipeline into 𝟰𝘅 𝗳𝗮𝘀𝘁𝗲𝗿 embeddings and 𝟲𝘅 𝗳𝗮𝘀𝘁𝗲𝗿 chunking optimization is a different skill
27
594
Hardest part of building a gud RAG system? ZIP - preserve structure PDF/IMGS - ocr when needed Webpages - clean extraction YT - transcript SRT/VTT - timestamp aware chunks Every source needs its own extraction+chunking+rejecting strategies I learned more today than I expected
2
1
19
551
So, I'm working on something Similar to NotebookLM. And I call it DOX Here's current update regarding it Join the waitlist : dox.sud-o.app #chaicode #chaicohort @Hiteshdotcom @ChaiCodeHQ @piyushgarg_dev Stay tuned !
1
16
616