Architecting self-evolving AI systems at the intersection of Evolution Core, Governance & Verification, Mathematical Framing, and System Architecture.
I build technology from bare-metal scratch — no framework shortcuts, no encapsulated black boxes. All implementations are hand-coded, verifiable, and open-sourced permanently.
Most engineers build on existing stacks.
I reconstruct the underlying stack itself: from mathematical theory, hardware circuits, and low-level system firmware, up to LLM underlying architecture, CUDA kernel optimization, and verifiable self-evolving agent systems.
This repository archive records continuous bottom-up engineering and AI paradigm exploration. Every project is:
- ✅ Runnable
- ✅ Verifiable
- ✅ Serves as practical prior art for underlying system and intelligent infrastructure research
"Open source is not a hobby, but a public technical manifesto of full-stack underlying engineering."
Long live ❤️ Open Source.
Full coverage of the entire computer technology stack, 100% pure C implemented from scratch. Rigorous bottom-up construction to verify the underlying core principles of all hardware, system, network, business and intelligent technologies.
| Repository | Description | Reference Course & Materials | Link |
|---|---|---|---|
mini-math-theory |
University-level mathematics & computer theory for system and AI modeling | MIT 6.006/6.046J, 6.042J, 6.045J, 18.05/18.06/18.065, 6.441; Stanford CS229; Sipser Computation Theory | 🔗 View |
mini-hardware-physical |
Bottom-up hardware design and physical circuit implementation | MIT 6.004/6.175/6.823/6.5900/6.5930/6.5950; CMU 18-447/15-418; Stanford CS144/CS149/EE282; UC Berkeley CS261 | 🔗 View |
mini-firmware-boot |
Lightweight bare-metal firmware and bootloader program development | UEFI PI Spec, TianoCore EDK II, GRUB2, Das U-Boot, TPM 2.0 Spec, TCG PC Client, NIST SP 800-193, Intel/ARM Trusted Firmware | 🔗 View |
mini-os-driver-sys |
Handwritten OS kernel, driver and virtualization underlying logic | MIT 6.828 (xv6), CMU 15-410; Intel VT-x/AMD-V; OCI Runtime Spec; CS:APP; Linux Kernel LSM | 🔗 View |
mini-lang-compiler-vm |
From-scratch programming language, compiler & virtual machine implementation | Stanford CS143/CS242; CMU 15-745; MIT 6.945; Dragon Book; Modern Compiler Book; MLIR/TVM/XLA | 🔗 View |
mini-network-dist-proto |
Distributed network architecture and core communication protocol from scratch | MIT 6.824/6.829; Stanford CS144; CMU 15-721; Raft/Paxos Papers; DDIA; IETF RFC Standards | 🔗 View |
mini-data-store-search-vec |
Vector database underlying storage, indexing and similarity retrieval implementation | CMU 15-445/645; MIT 6.830; Stanford CS245; FAISS/Milvus/Annoy; LevelDB/RocksDB; Lucene | 🔗 View |
mini-data-engine-lakehouse |
Self-built data computing engine and lakehouse architecture underlying system | Kimball DWH Toolkit, Delta Lake/Iceberg/Hudi Spec, Spark/Flink/Kafka, ClickHouse/DuckDB OLAP Theory | 🔗 View |
mini-backend-api-business |
Backend service architecture, API interface and business logic bottom-up construction | OAuth2 RFC 6749, JWT RFC 7519; DDD/CQRS/Event Sourcing; REST/GraphQL Spec | 🔗 View |
mini-frontend-client-web |
Native web frontend rendering, interactive logic and client principle implementation | W3C Specs, Chromium Blink, V8 Engine, WhatWG Fetch Standard | 🔗 View |
mini-graphics-render-game |
Spatial computing, rendering engine & game physics loop implementation | MIT 6.837, OpenGL 4.6/Vulkan 1.3, OpenXR/WebXR, ECS Game Architecture | 🔗 View |
mini-media-av-rtc |
Audio and video processing, real-time RTC transmission and media service underlying logic | H.264/AVC, ITU-T.81, WebRTC 1.0, HLS/DASH RFC, FFmpeg Architecture | 🔗 View |
mini-cloud-native-sre |
Cloud-native architecture, service orchestration and SRE stability governance from scratch | Kubernetes/Borg, Istio/Envoy, OpenTelemetry, Google SRE Book, Brendan Gregg Performance Theory | 🔗 View |
mini-security-crypto-web3 |
Underlying cryptographic algorithms, network security and Web3 core protocol implementation | NIST FIPS, MIT 6.858, OWASP Top10, zk-SNARKs, Intel SGX/AMD SEV, TCG Standards | 🔗 View |
mini-ai-ml-intelligent |
From-scratch implementation of underlying machine learning and intelligent algorithm frameworks | Stanford CS229, MIT 6.036; PyTorch/TensorFlow; vLLM/TensorRT-LLM; CLIP/Stable Diffusion | 🔗 View |
mini-iot-robot-edge |
Edge computing, IoT terminal and robot underlying control system development | ARM Cortex-M TRM, FreeRTOS, ROS2, TinyML, IEC 61131-3, ARM TrustZone-M | 🔗 View |
mini-hpc-sci-compute |
High-performance parallel computing and scientific numerical simulation implementation | MIT 6.172, Stanford CS149; CUDA/OpenMP/MPI; BLAS/LAPACK; Roofline Model | 🔗 View |
mini-eda-fpga-asic |
EDA tool development, FPGA logic & ASIC chip underlying design | IEEE 1364/1800 Verilog, UVM, RISC-V ISA, Synopsys/Cadence EDA Tools, NoC Theory | 🔗 View |
mini-software-eng-product |
Bottom-up standardized software engineering system and project practice | C4 Model, Conventional Commits, Clean Code, SAFe/Scrum, SonarQube Testing Standards | 🔗 View |
mini-app-industry-product |
Industrial-grade embedded application development and engineering practice | Enterprise ERP/CRM Standards, Industrial Embedded Specs, FinTech/HealthTech Engineering Norms | 🔗 View |
Focus on large model architecture inference, kernel optimization, automated research, and multi-agent engineering workflows.
| Repository | Description | Link |
|---|---|---|
Apex |
Fully self-developed LLM native reasoning architecture | 🔗 View |
OpenClaude-Mythos |
Recurrent depth Transformer model structural inference & parsing | 🔗 View |
OpenGemini3.1-Pro |
Layer-by-layer analysis of native multimodal sparse MoE architecture | 🔗 View |
OpenChatgpt5.5 |
In-depth parsing of dense Transformer + sparse MoE hybrid backbone | 🔗 View |
| Repository | Description | Link |
|---|---|---|
DeepSci |
End-to-end automated full-cycle scientific research tool | 🔗 View |
PRForge |
Multi-agent autonomous code review & engineering workflow system | 🔗 View |
BTCBoard |
Multi-agent adversarial verification market analysis platform | 🔗 View |
HoTT-CatWorld |
HoTT & higher category theory based verifiable AI reasoning system | 🔗 View |
Trace2Train |
Convert program execution traces into model training data & checkpoints | 🔗 View |
LLMSched |
LLM serving KV cache, token & batch scheduling simulation | 🔗 View |
KernelLab |
Handcrafted CUDA kernels for LLM core computing hot paths | 🔗 View |
IRPlanner |
LLM computation graph compilation & execution plan optimization | 🔗 View |
Bottom-up underlying runtime and resource scheduling system for scalable autonomous AI.
| Repository | Description | Link |
|---|---|---|
OntoLoop 🌟 |
Scalable verifiable autonomous AI runtime with swarm intelligence & self-evolution capability | 🔗 View |
Apeinx |
AI underlying resource scheduling system, "Linux for AI" (GPU/Token/Model/Agent unified management) | 🔗 View |
Triton-Agent |
Autonomous GPU kernel construction, tuning & optimization agent | 🔗 View |
Languages: C · Rust · Python · CUDA Core Domains: Mathematical Formal Verification · Bare-metal Hardware & Kernel Compiler & VM · LLM Architecture Reverse Engineering CUDA Kernel Optimization · Multi-Agent Autonomous Systems AI Infrastructure Scheduling · Automated Scientific Research Philosophy: Verifiable · Reproducible · From Scratch · Open by Default
- 🔍 Explore: Browse any repository above — all code is open for inspection and contribution
- 🤝 Contribute: PRs, issues, and technical discussions are warmly welcomed
- 💡 Collaborate: Interested in bottom-up AI infrastructure research? Let's connect
"The best way to predict the future is to implement it from first principles."
Built with ❤️ and bare-metal determination by rootkiller6788
🔐 All projects are open source under MIT License unless otherwise specified