htsh

Joined 17 years ago

Karma

891

Submissions

296

Member since

May 15

Applied LLM Engineer working on games and full stack apps.

Formerly engineer @ a large pharma focused on data science projects, primarily JS and Python, also interested in copyright / open source licensing.

hitesh at gmail.

View on Hacker News

Recent submissions

Comment · 4 months ago

curious, how does one configure something like this for AWS Lambda? Appreciate it.

I am assuming it is a bunch of manual work.

Comment · 4 months ago

Is it safe to assume they will come down in 2027?

Comment · 5 months ago

i much prefer svelte but the rest of the world does not

Comment · 8 months ago

yes! especially b/c i want to process a lot of email and directories full of old, personal documents

Comment · 8 months ago

are we sure the RAM market will stop being insane in a year or two or could this be the new norm?

Comment · 9 months ago

thanks! came in here to ask this.

we can do much better with a cheap model on openrouter (glm 4.7, kimi, etc.) than anything that I can run on my lowly 3090 :)

Comment · 1 year ago

I have been doing this with claude code and openai codex and/or cline. One of the three takes the first pass (usually claude code, sometimes codex), then I will have cline / gemini 2.5 do a "code review" and offer suggestions for fixes before it applies them.

Comment · 1 year ago

curious, why the 30b MoE over the 32b dense for local coding?

I do not know much about the benchmarks but the two coding ones look similar.

Comment · 1 year ago

A lot of us have ryzen / nvidia combos... hopefully, soon, though.

Comment · 1 year ago

assuming you want to run entirely in GPU, with 12gb vram, your sweet spot is likely the distill 14b qwen at a 4bit quant. so just run:

ollama run deepseek-r1:14b

generally, if the model file size < your vram, it is gonna run well. this file is 9gb.

if you don't mind slower generation, you can run models that fit within your vram + ram, and ollama will handle that offloading of layers for you.

so the 32b should run on your system, but it is gonna be much slower as it will be using GPU + CPU.

prob of interest: https://simonwillison.net/2025/Jan/20/deepseek-r1/

-h