Drivers & Communication
From STM32WLE interface drivers, FatFS storage, RTC timestamps, and one-wire acquisition to instrument-based debugging.
View Resume →Embedded Software · Embedded Linux · Engineering Review
Embedded Personal Engineering Practice
I am Xiong Xiao, an Internet of Things undergraduate at Chongqing University of Science and Technology. My job direction is embedded software development and embedded Linux development. My projects cover low-level drivers, communication stability, Linux multithreaded applications, hardware-software debugging, and also extend to AI Agent Tools, data-processing pipelines, and engineering utilities. This site records how I locate problems, verify solutions, and turn ideas into runnable systems.
6 Portfolio Projects
1 CATARC Internship
16 Awards and IP Results
Capability Path
Three Engineering Entries
Real problems: locate, verify, deliver, review.
Linux Concurrency
Embedded Linux threading, blocking AI request isolation, Framebuffer display, and network fallback paths for avoiding system freeze.
Browse Projects →Engineering Review
Writing projects as reusable technical records: context, diagnosis, verification, results, and methods that can transfer to the next system.
Read Blog →Featured Projects
Representative Projects
CATARC Multi-Interface Embedded Development and Debugging
Problem：When several communication interfaces and acquisition paths were developed in parallel, driver reuse was weak and application symptoms could hide lower-level interface faults.
Technical Action：Separated HAL, BSP, and App layers, wrapped I2C, UART, SPI, RS485, one-wire, and storage modules, then used logic-analyzer and oscilloscope traces to locate level-matching errors.
Outcome：Improved code reuse by more than 50%, and the debugging notes plus compatibility recommendations were absorbed into the team interface standard.
STM32WLECFatFSRTCDriver Debugging
Read Technical Review → Smart Life Emotional Companion System
Problem：AI requests, audio playback, sensor sampling, and display refresh shared limited system resources, so one blocking call could become whole-system lag.
Technical Action：Split sampling, interaction, and rendering with pthread, integrated ALSA, OpenSSL, and MPlayer by cross-compilation, and designed a network-failure fallback path.
Outcome：Decoupled AI interaction from environment sampling, keeping local sampling, safety alerts, and basic feedback available during unstable network conditions.
LinuxGEC6818pthreadFramebufferDeepSeek API
Read Technical Review → 3D Machine Vision Chinese Chess Robot
Problem：Vision recognition results could not directly drive the robotic arm because image coordinates, board coordinates, and motion coordinates carried different error sources.
Technical Action：Connected YOLOv5 detection, nine-point board calibration, Alpha-Beta search, and PWM robotic-arm control, then added error compensation across the execution loop.
Outcome：Reached +/-0.5 mm repeat positioning accuracy and inference response within 300 ms; the project received national innovation-program approval and an invention patent.
YOLOv5Alpha-BetaPWMRobotic ArmInnovation Project
Read Technical Review → 
Data Knowledge Insight Demo
Problem：Raw CSV files and medical text are hard to query, analyze, and reuse directly as knowledge assets.
Technical Action：Created an end-to-end pipeline for data cleaning, text chunking, entity-relation extraction, knowledge graph output, KGQA, statistical analysis, NL2SQL evaluation, and FastAPI interfaces.
Outcome：Generated cleaned data, medical knowledge graph files in JSON/GraphML/HTML, QA examples, analysis reports, charts, and an API demo.
PythonFastAPIKnowledge GraphAgentData AnalysisNL2SQL
Read Technical Review → View GitHub Repository → Latest Notes
Latest Blog Posts
From an Open Source Competition to Knowledge Graph QA: Building the Data Knowledge Insight Agent Pipeline
A review of the data-knowledge-insight-demo project: organizing cleaning, knowledge graphs, KGQA, NL2SQL, and FastAPI into an end-to-end agent demo, adapted for DataMate and Nexent, awarded Third Prize in the ModelEngine Open Source Project Contribution Competition and merged into AgentsHub.
PythonFastAPIKnowledge GraphAI AgentNL2SQL
How Multithreading Helped Avoid System Freeze in an Embedded Linux Project
A review of concurrency decomposition in an embedded Linux multimodal system: to avoid blocking AI requests and UI lag, I isolated sampling, interaction, and display paths.
Embedded LinuxpthreadConcurrencySystem Debugging