ml infra @nvidia (2x), prev @slackhq & @figma, cs @ ut austin ‘27 love learning & all views are my own

localhost
Pinned Tweet
there are two phases to every internship: 1. eating all of the protein snacks in the office 2. stomach problems
1
12
1,404
if you do it right, you only need to do technical interview prep one time
31
unironically, i mostly learned how to code through leetcode at first without it, my systems classes would have been much much rougher to go through
4
70
the tradeoffs surrounding using a wide column database are so interesting
1
2
71
in short, it makes data locality easier to reason about at the cost of flexibly being able to represent every kind of relationship.
32
everything is dsa everything is a data structure or an algorithm
There is only one thing that is tough to master in tech, and that is DSA. Everything else, from system design to machine coding, is just a joke compared to it. Focus on DSA if you are starting college. All other skills are much easier to learn.
1
129
instead of feeling the need to turn every problem i have into a startup idea, often detracting away from solving the original problem, i’ve loved using ai to create personal software some examples: - a site for learning how to sing usetoner.com (the only publicly accessible one haha) - an app where i can upload contacts from my phone, and keep track of the last time i reached out to selected contacts to help me keep in touch - personal email agent (self explanatory haha) - (currently in progress) my own cloud agent control plane with persistent sandboxes, so that i can code from my laptop, my phone, with my voice, from anywhere. - a to do list app to help me keep track of the growing stack of things in my life, with integrations to my calendar & email hopefully soon! doing these for myself have been huge productivity unlocks :)
4
5
329
but overall, instead of waiting for an existing platform to do what i want, i can just go and do the thing for myself don’t like a particular user interaction? do it my way don’t like a particular experience or feature? do it my way don’t like waiting for an integration? do it my way so on and so forth
30
16 pages later… lol
if there’s anything that is difficult to estimate, it’s how long it takes to write a design document
1
3
258
i wonder what etched thinks about jev
1
2
196
this is a really interesting discussion for people like myself, who haven’t had a single professional experience without the presence of llms. writing code use to be a problem in the way of solving problems. for people who enjoyed writing code, this was awesome. but i’ve really enjoyed how i’m able to focus more on solving problems WITH code very interesting times
My software engineer friends say AI has made their jobs mostly grunt work. The models automate the parts they used to enjoy, leaving humans with the hardest, least enjoyable parts of the job.
1
193
if there’s anything that is difficult to estimate, it’s how long it takes to write a design document
2
4
426
shooting into the abyss @openai re: the gmail connector - if multiple emails are connected to an account, the UI doesn’t properly show which account the email came from - replies don’t draft the reply to the original thread it’s in response to 🙏🏾
92
there is nothing difficult to understand. only things that are unfamiliar
1
2
98
as promised, notes on the units of compute are live :) was supposed to do a homework assignment, but got caught up and spent a lot of time the past couple days on this after taking a closer look into the work @modal is doing in this area. still have some visuals to add, but core content is essentially done. tezock.com/bookshelf
the units of compute: thread process container *microvm virtual machine physical machine
1
4
711
planning on spending some time diving into: - computer architecture - gpus - machine learning - reviewing operating systems / file systems - observability for the immediate future after this, but i feel like im finally hitting that point of, for the most part, understanding a lot of the core design decisions i sought learning when i first got interested in infra. onwards and upwards!
1
1
62
welp. now to do my machine learning assignment.
2
22
the units of compute: thread process container *microvm virtual machine physical machine
2
9
1,556
notes soon
1
90
one of my favorite parts about software engineering is that no tool is right all the time. there is only the right tool for the job.
1
3
210