██████╗ █████╗ ██████╗ █████╗ ████████╗ ██████╗ ██████╗ ██╗ ██╗███████╗
██╔══██╗██╔══██╗██╔══██╗██╔══██╗╚══██╔══╝██╔═══██╗██╔═══██╗██║ ██║██╔════╝
██║ ██║███████║██║ ██║███████║ ██║ ██║ ██║██║ ██║██║ ██║███████╗
██║ ██║██╔══██║██║ ██║██╔══██║ ██║ ██║ ██║██║ ██║██║ ██║╚════██║
██████╔╝██║ ██║██████╔╝██║ ██║ ██║ ╚██████╔╝╚██████╔╝╚██████╔╝███████║
╚═════╝ ╚═╝ ╚═╝╚═════╝ ╚═╝ ╚═╝ ╚═╝ ╚═════╝ ╚═════╝ ╚═════╝ ╚══════╝
296,000+ lines of real code. 25+ components. 5 languages. All local. No cloud.
CODA is a local-first AI operating system. No API keys, no subscriptions, no telemetry. Everything runs on your hardware. I built it alone.
+-----------+
| HUMAN |
+-----+-----+
|
+----------+----------+
| | |
v v v
+--------+ +--------+ +--------+
| CODA | | KARNA | | CODA |
| OS | | (2.2B) | | Forge |
| 5 UIs | | Voice | | 3D |
+---+----+ +---+----+ +---+----+
| | |
+-----+----+----+-----+
| |
v v
+-----------+ +-----------+
| BHAASM | | Cloud |
| Brain | | GPU |
| 43K loc | | Training |
+-----------+ +-----------+
|
43,768 lines | Python From-scratch 2.2B decoder-only transformer. RoPE + GQA + SwiGLU + MoE (8 experts). LoRA fine-tuning, debate engine, generation refinement. |
32,800 lines | Python 17-stage photorealistic 3D human reconstruction. Phone rotation video -> textured, watertight mesh. COLMAP + ECON + PIFuHD + Gaussian Splatting. |
|
~3,000 lines | Python No-GPU 3D body scanning. CPU-only, zero neural networks. Silhouette widths at 360 angles -> ellipse fit -> mesh. Works on any laptop. |
1,785 lines | Python Active defense through deception. Fake services, canary tokens, honeypots. MITRE ATT&CK mapped. 12 techniques. |
|
1,204 lines | Python Declarative DSL for video generation. Write scenes in text, render with FFmpeg. Plugin-based, sandboxed rendering. |
~800 lines | Python AES-256-GCM, Argon2, PBKDF2, MFA. Tamper-evident audit chains. Production-ready. No dependencies. |
| Domain | Tools |
|---|---|
| AI/ML | PyTorch, HuggingFace, LoRA, MoE, BPE Tokenizer |
| 3D | COLMAP, ECON, PIFuHD, Gaussian Splatting, TripoSR |
| Generation | Stable Diffusion XL, AnimateDiff, I2V, NIM FLUX |
| Voice | Silero VAD, faster-whisper, edge-tts, MediaPipe |
| Web | FastAPI, Next.js 16, React 19, TailwindCSS |
| Desktop | PySide6, Electron 28, Flutter/Dart |
| Security | AES-256-GCM, PBKDF2, MFA, Audit Chain |
| Cloud | Kaggle T4, Modal T4, Colab T4, Lightning T4 |
10 composable projects forming a virtual device:
| Project | Purpose |
|---|---|
| Cosmic Explore | 3D codebase intelligence |
| Cosmic Voice | Local voice pipeline (STT + TTS) |
| Cosmic Vision | Gesture recognition + hand tracking |
| Cosmic Canvas | Sketch-to-3D generation |
| Cosmic Core | AI operating system shell |
| Cosmic Flow | Visual pipeline builder |
| Cosmic Atlas | 3D code navigation |
| Cosmic Debug | Visual debugger |
| Cosmic Connect | Service registry + discovery |
| Cosmic Shield | Supply chain security |
296,000+ lines of code
30 public repositories
25+ working components
5 programming languages
4 cloud GPU tiers
3 UI frameworks
1 developer
- Local-first -- No cloud. Your hardware, your data.
- Composable -- Each piece works standalone.
- Agent-driven -- AGNI creates, VEDA validates, VIJAY interacts.
- Self-improving -- TRISHUL loop: generate, evaluate, improve.
- Multi-accelerator -- CPU + GPU + iGPU + NPU.
Built with curiosity and a laptop. No team. No funding. Just code.