Skip to content
View mohi-devhub's full-sized avatar
🎯
Grinding
🎯
Grinding

Organizations

@ruxailab

Block or report mohi-devhub

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 don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
mohi-devhub/README.md

I work on LLM inference, evaluation and post-training, mostly measuring how serving systems actually behave and building tools around what I find.

Portfolio · LinkedIn · X · ORCID

Research

  • Reload or Recompute? (preprint): a break-even rule for when a resumed conversation should reload its KV cache instead of recomputing it, tested on L4, A10 and A10G with predictions frozen in advance.
  • Leaderboard Noise: "deterministic" greedy decoding changed 2.0–4.6% of answers across GPUs and reversed a near-tied ranking.
  • Grammar vs Bits (preprint): strict grammar decoding gave no clear recovery of what 4-bit quantization loses in tool calling; the output format mattered far more.

Post-training

  • Pixel Miser (write-up): post-trained Inkling 2 Small on Tinker to stop over-thinking. Output tokens fell from 3,109 to 74 and finish rate rose from 49% to 99%.
  • Nightly (in progress): a coding agent that retrains on its own test-verified attempts each night, with hidden-test graders red-teamed against cheating.

Projects

  • Kestrel: multi-tenant GPU compute platform on Kubernetes with quotas, GPU-second metering and autoscaling.
  • kea: real-time incident root-cause engine. 1st place, OpenAI Codex Community Hackathon.
  • AntiVibe: Claude Code skill that explains AI-written code. 1,100+ stars, Mintlify OSS Program.
  • SentinelLM: LLM proxy with a concurrent safety-evaluation pipeline.

Experience

  • Google Summer of Code 2026, RUXAILAB: built the Focus Group Management and Analysis Module (24 merged PRs).
  • Koncepts: Product Developer Intern, 165+ PRs to an AI EdTech platform serving 200K+ learners.
  • India Builds with Claude: 1 of 6 presenters selected from 2,300+ applicants.

Stack

Python PyTorch vLLM Tinker Modal TypeScript FastAPI PostgreSQL Docker Kubernetes

Pinned Loading

  1. antivibe antivibe Public

    Learn what AI writes, not just accept it. A Claude Code skill that turns AI-generated code into educational deep dives.

    Shell 1.1k 83

  2. SentinelLM SentinelLM Public

    A FastAPI middleware that intercepts LLM requests and responses in real time, scoring them for prompt injection, PII leakage, toxicity, hallucination, and relevance.

    Python 5

  3. voidwalker voidwalker Public

    A browser extension + local MCP server that streams localStorage, IndexedDB, cookies, and DOM mutations to Claude, Cursor, and any MCP-compatible AI client.

    TypeScript 2 3

  4. StudyQuest StudyQuest Public

    A terminal-style, AI-powered learning platform with adaptive quizzes, real-time XP tracking, and a minimalist B/W dashboard.

    Python 1