Skip to content
View haiyang5535's full-sized avatar
5 followers · 3 following

Achievements

Achievement: Pair Extraordinairex4Achievement: Pull Sharkx3Achievement: YOLOAchievement: Quickdraw

Achievements

Achievement: Pair Extraordinairex4Achievement: Pull Sharkx3Achievement: YOLOAchievement: Quickdraw

Highlights

  • Pro

Block or report haiyang5535

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.

Maximum 250 characters. Please don’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 user’s behavior. Learn more about reporting abuse.

Report abuse
haiyang5535/README.md

Haiyang Wu

Edge AI & Computer Vision engineer. Math-stat background, self-taught into ML systems.

Currently building SMBoost — a reliability harness for small LLMs at the edge. Came out of a failure mode I observed running a 3B VLM on Jetson Orin during the Machina AI pilot.

Previously founded Machina AI — zero-cloud visual inspection for high-mix manufacturing. Built EdgeRunner, signed a pilot LOI with a Shenzhen powder metallurgy manufacturer, learned that ToB sales on a single-founder timeline is a different problem than the engineering.

Pinned

→ edge-runner — Real-time vision system on NVIDIA Jetson. YOLO + VLM dual-thread architecture. <12ms P99 latency.

→ smboost — Reliability harness for 2B–7B local LLMs. State machine + GBNF constraints + adaptive scoring.

Stack

Python · NVIDIA Jetson · TensorRT · YOLO · VLMs · LangGraph · llama.cpp · Computer Vision · ML Systems · Edge Deployment

Contact

[email protected] · haiyang5535.github.io · linkedin.com/in/h-wu

Pinned Loading

  1. edge-runner edge-runner Public

    Real-Time Edge AI Vision System — YOLO + VLM on NVIDIA Jetson Orin Nano | Machina AI

    Python 1

  2. smboost smboost Public

    Decoding-time harness that lifts small open-weight LLMs (Qwen 2.5 2B Q4) on verifier-friendly benchmarks via parallel self-consistency, per-sample program verifiers, and raw-anchored majority voting.

    Python 1