Skip to content
View rhgo1749's full-sized avatar
๐Ÿค”
๐Ÿค”
4 followers · 5 following

Achievements

Achievement: Pull Sharkx3Achievement: QuickdrawAchievement: Arctic Code Vault Contributor

Achievements

Achievement: Pull Sharkx3Achievement: QuickdrawAchievement: Arctic Code Vault Contributor

Block or report rhgo1749

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please donโ€™t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this userโ€™s behavior. Learn more about reporting abuse.

Report abuse
rhgo1749/README.md

H4v3Lab

Building practical AI systems โ€” local LLM inference, agent workflows, and Android tools.

๋กœ์ปฌ LLM ์ถ”๋ก ๊ณผ ์—์ด์ „ํŠธ ์ž๋™ํ™”๋ฅผ ์ง์ ‘ ๊ตฌ์ถ•ํ•˜๊ณ , ์„ฑ๋Šฅ๊ณผ ์‹ ๋ขฐ์„ฑ์„ ์‹ค์ธกํ•˜๋ฉฐ ๊ฐœ์„ ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.

What I work on

Area What I build & investigate
LLM inference & serving Quantization, heterogeneous GPU serving, context/KV-cache behavior, and performance benchmarking
Agent workflows Hermes-based automation, GitHub โ†” Kanban lifecycle integration, review and recovery reliability
Android CtrlHangul, a free Korean/English input-method app (source repository is private)
Engineering practice Reproducible experiments, regression tests, evidence-backed performance and reliability work

Open-source contribution

  • Strata โ€” contributed --shared-expert-arena upstream at the maintainer's request. See PR #129 and v0.1.30.

About / ์†Œ๊ฐœ

English

I build and experiment with LLM inference systems, heterogeneous serving, and efficient local AI runtimes. I also work on automation and Android tools, with a focus on measured results and maintainable systems.

Open to full-time roles, research collaborations, and graduate research opportunities.

ํ•œ๊ตญ์–ด

LLM ์ถ”๋ก  ์‹œ์Šคํ…œ, ์ด๊ธฐ์ข… GPU ์„œ๋น™, ํšจ์œจ์ ์ธ ๋กœ์ปฌ AI ๋Ÿฐํƒ€์ž„์„ ๊ฐœ๋ฐœํ•˜๊ณ  ์‹คํ—˜ํ•ฉ๋‹ˆ๋‹ค. Hermes ๊ธฐ๋ฐ˜ ์ž‘์—… ์ž๋™ํ™”์™€ Android ์ž…๋ ฅ๊ธฐ ๊ฐœ๋ฐœ๋„ ์ง„ํ–‰ํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ์‹ค์ธก ๊ฒฐ๊ณผ์™€ ๊ฒ€์ฆ ๊ฐ€๋Šฅํ•œ ๊ฐœ์„ ์„ ์ค‘์š”ํ•˜๊ฒŒ ์ƒ๊ฐํ•ฉ๋‹ˆ๋‹ค.

์ •๊ทœ์ง ์ฑ„์šฉ, ์—ฐ๊ตฌ ํ˜‘์—…, ๋Œ€ํ•™์› ์—ฐ๊ตฌ ๊ธฐํšŒ์— ์—ด๋ ค ์žˆ์Šต๋‹ˆ๋‹ค.

ไธญๆ–‡

ๆˆ‘ไธ“ๆณจไบŽ LLM ๆŽจ็†็ณป็ปŸใ€ๅผ‚ๆž„ๆŽจ็†ๆœๅŠกไปฅๅŠ้ซ˜ๆ•ˆๆœฌๅœฐ AI ่ฟ่กŒๆ—ถ็š„ๅผ€ๅ‘ไธŽๅฎž้ชŒ๏ผŒไนŸๆŽข็ดขๆ™บ่ƒฝไฝ“่‡ชๅŠจๅŒ–ๅ’Œ Android ๅทฅๅ…ทๅผ€ๅ‘ใ€‚

ๆฌข่ฟŽๅ…จ่Œๅทฅไฝœๆœบไผšใ€็ ”็ฉถๅˆไฝœไปฅๅŠ็ ”็ฉถ็”Ÿ็ ”็ฉถๆœบไผšใ€‚

ๆ—ฅๆœฌ่ชž

LLM ๆŽจ่ซ–ใ‚ทใ‚นใƒ†ใƒ ใ€็•ฐ็จฎใƒใƒผใƒ‰ใ‚ฆใ‚งใ‚ขใ‚’ๆดป็”จใ—ใŸใ‚ตใƒผใƒ“ใƒณใ‚ฐใ€ๅŠน็Ž‡็š„ใชใƒญใƒผใ‚ซใƒซ AI ใƒฉใƒณใ‚ฟใ‚คใƒ ใฎ้–‹็™บใƒปๆคœ่จผใซๅ–ใ‚Š็ต„ใ‚“ใงใ„ใพใ™ใ€‚ใ‚จใƒผใ‚ธใ‚งใƒณใƒˆ่‡ชๅ‹•ๅŒ–ใจ Android ใƒ„ใƒผใƒซใ‚‚้–‹็™บใ—ใฆใ„ใพใ™ใ€‚

ใƒ•ใƒซใ‚ฟใ‚คใƒ ใฎๆŽก็”จใ€็ ”็ฉถๅ”ๅŠ›ใ€ๅคงๅญฆ้™ขใงใฎ็ ”็ฉถๆฉŸไผšใ‚’ๆญ“่ฟŽใ—ใฆใ„ใพใ™ใ€‚

Contact

โ˜• Optional support for independent development: Buy Me a Coffee. Support is voluntary and does not unlock features or benefits.

Pinned Loading

  1. qwen3.8-flash-next-exllamav3-3x5070ti-recipe qwen3.8-flash-next-exllamav3-3x5070ti-recipe Public

    Reproducible ExLlamaV3 recipe and optimization log for Qwen3.8-Flash-Next on 3x RTX 5070 Ti 16GB

    Python

  2. qwen3.8-flash-next-strata-gpu-per-lane-recipe qwen3.8-flash-next-strata-gpu-per-lane-recipe Public

    Measured 3x RTX 5070 Ti recipe for Qwen3.8-Flash-Next on the Strata multi-GPU fork: 3x262K lanes, shared expert arena, and real Hermes concurrency results.

    Python

  3. Niko1221/Strata Niko1221/Strata Public

    Qwen3.8-Flash-Next on any consumer hardware: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input.

    C++ 27.1k 2.8k

  4. Strata-Lanes Strata-Lanes Public

    Forked from Niko1221/Strata

    Qwen3.8-Flash-Next (125B MoE) on a 12-24 GB NVIDIA GPU + 64 GB RAM: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input.

    C++