Skip to content
View Alexander390370's full-sized avatar
3 followers · 1 following

Block or report Alexander390370

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
Alexander390370/README.md

👋 Hi there, I'm Alexander

21-year-old developer in China, working at the intersection of AIoT, edge AI, and embedded systems — and lately, at the boundary between AI and quantitative research.

Profile Views Focus Hardware

🛠️ What I'm Working With

Area Stack
Hardware BGA rework, motherboard repair, edge device integration
IoT / Embedded ESP32 / ESP32-S3 / ESP32-CAM, MQTT, multi-sensor fusion
AI / Edge Ollama, llama.cpp, YOLOv8, Whisper, local LLM deployment
Agent Infra DeepSeek Harness plugins, MCP servers, multi-model orchestration
Full-Stack WSL2, Docker, FastAPI, Python 3.12

🚀 Featured Projects

🔬 Quant Research

Repo What it is Outcome
ashare-quant-research Full engineering log: A-share screening + Qlib factor backtest + mutation-tested validation 9 strategies validated, 0 survived — a negative-result study

🧠 Agent Infrastructure

Repo What it is Key number
dsh-multi-model-orchestration Route tasks to Kimi / Qwen / GLM / DeepSeek by strength, with config self-check 30 KB, pure local
dsh-laya-mcp Typed-decision model as an MCP server for DeepSeek Harness ~55 ms/item, zero tokens

🖥️ Local LLM on Consumer GPU

Repo What it is Key number
Bonsai 27B on 8GB VRAM Ternary 27B at 16K–64K context on RTX 4060 18.8–20.5 tok/s
Qwen3.5 4B on 8GB VRAM Two 4B models at 64K context + "is uncensored really uncensored?" test 4.3 GB, 47–60 tok/s

🏭 Data & Generation Pipelines

Repo What it is Key number
OmniForge Full-modal private annotation (text/image/audio/video) + local AI pre-labeling 20–30 person teams
AIGC Local Image Pipeline Batch generation + API client on Stable Diffusion Forge Fully local
SD Forge on 8GB VRAM Deployment log for Forge on constrained hardware CUDA 12.1 + Torch 2.12

🔌 Embedded

Repo What it is
ESP32 Edge AI Security Two-node (S3 + CAM) sensor gateway + local 4B multimodal judgement → TTS + WeChat
ESP32 PWM Fan Controller DC motor control with transistor switch, Web UI, and OLED status

📌 What I'm Focused On Right Now

  • OmniForge — AI-assisted data annotation for startup clients
  • DeepSeek Harness plugins — making multi-model orchestration actually reliable
  • Squeezing big models into small VRAM — 27B on 8 GB is a solved problem now
  • Honest negative results — the quant repo documents 9 dead strategies, because "this doesn't work" is worth publishing too

💡 A Principle I Work By

Document what actually happened, not what was supposed to happen.

Every repo here includes the traps, the silent failures, and the measured numbers — because the next person's afternoon is worth more than my vanity.

📫 How to Reach Me

"Bringing AI from the cloud to the edge, one device at a time."

Pinned Loading

  1. OmniForge-Data-Annotation OmniForge-Data-Annotation Public

    Full-modal private data annotation platform with local AI pre-annotation (Ollama, YOLO, Whisper) and human-in-the-loop.

    3

  2. Bonsai-27B-8GB-VRAM-Setup Bonsai-27B-8GB-VRAM-Setup Public

    Running Ternary Bonsai 2 27B on 8GB VRAM with 64K context

    1

  3. esp32-edge-ai-security esp32-edge-ai-security Public

    Python 2

  4. sd-forge-8gb-vram-setup sd-forge-8gb-vram-setup Public

    Running Stable Diffusion Forge with uncensored community models on an 8GB VRAM laptop via WSL2.

    2

  5. aigc-local-image-pipeline aigc-local-image-pipeline Public

    A fully local AIGC image generation system built on Stable Diffusion Forge: batch generation, reusable API client, Windows one-click launchers.

    Python 2

  6. esp32-pwm-fan-controller esp32-pwm-fan-controller Public

    ESP32-based PWM speed controller for a DC motor, with an S8050 low-side switch and an OLED readout.

    C++ 2