Lead Autonomous Systems & Punjabi NLP Architect
Building the future of local, private, and self-evolving multi-agent intelligence.
I am a software engineer and AI researcher specializing in autonomous agent architecture, self-evolving codebases, and Gurmukhi NLP. My work focus is on pushing the boundaries of private local execution on Apple Silicon (M-series GPUs), implementing multi-agent swarms that research, plan, code, and self-debug in real time.
An autonomous, local multi-agent system comprising 19 specialized worker agents synced via an asynchronous event bus and DAG task scheduler.
- Self-Evolution Pipeline: Auto-scans python files, runs unit tests (121/121 passing), refactors modules, and applies updates in a secure sandbox.
- Multimodal Capabilities: Integrates Whisper-based voice (STT/TTS) and vision agents.
- Punjabi NLP: Locally fine-tuned Gurmukhi model adapters and ethical guardrails (Dharam, Sach, Daya).
An autonomous scientific discovery and medical research engine designed to synthesize hypotheses, simulate scenarios, parse arXiv & PubMed databases, and draft comprehensive papers.
An interactive AI-powered creative engine that creates music and generates beats.
☬ Sovereign Foundation Models (Nam-toon Studio)
Open-source Punjabi & Indic AI models and datasets on Hugging Face (1,300+ total downloads):
- 🧠 Sahaj-86M (Causal LM / State-Space Model): 86M parameter Gurmukhi/Punjabi foundation model trained on high-density reasoning, STEM CoT, and philosophy corpora with BPE tokenization.
- ⚡ Sahaj-30M (Edge & On-Device LM): 30.6M parameter byte-level causal LM achieving 0.149 Bits-Per-Byte (BPB) with sub-millisecond edge latency (~115MB RAM).
- 🎙️ Punjabi Studio TTS Voices: High-fidelity Piper / VITS ONNX neural voice synthesis models.
- 📚 Frontier Gurmukhi Datasets: Mahan Kosh Lexical Corpus, Punjabi STEM CoT, and Punjab Heritage Theological Corpus.
| Layer | Technologies & Tools |
|---|---|
| Languages | Python, JavaScript, Swift, Bash, Gurmukhi NLP |
| Agentic Frameworks | Swarm, CrewAI, Custom EventBus, DAG Task Scheduler |
| Local LLMs | Ollama, AirLLM (32B Escalation), local LoRA adapters |
| Vector & Databases | FAISS, SQLite, SQLite-VSS |
| Multimodal STT/TTS | Whisper, Custom TTS caching |
| Deployment & Ops | Docker, Makefile, GitHub Actions, Apple Silicon (Metal GPU) |
Connect & Collaborate:
GitHub · Hugging Face (Nam-toon Studio) · LinkedIn · gurpreetsingh5523@gmail.com
☬ ਵਾਹਿਗੁਰੂ ਜੀ ਕਾ ਖ਼ਾਲਸਾ ॥ ਵਾਹਿਗੁਰੂ ਜੀ ਕੀ ਫ਼ਤਹਿ ॥