"the future belongs to those who understand at a very deep level how to combine their unique expertise with what algorithms do best" — Pedro Domingos, The Master Algorithm
Independent Research & Systems.
Building local-first AI systems across reasoning, tool use, long-context agents, and multimodal workflows including text, image, audio, video-as-frames, voice notes, Telegram agents, Claude Code backends, and personal/household intelligence.
- Cogni-Brain Qwen Omni: Latest high-context Cogni-Brain iteration for DGX Spark: Qwen3.5-122B-A10B, a native multimodal model supporting text, image, and video inputs, served via spark-vLLM Docker. Compared with the earlier Nemotron brain, this variant expands local context from 131K to 262K, improves speed from ~24 to ~40 tok/s, and reaches 100/100 Tool-Eval.
- Cogni-Brain Omni Vision: Dedicated local multimodal perception agent for sandboxed DGX Spark workflows: Gemma 4 12B via vLLM with image input, audio/voice-note workflows, video-as-frames perception, multilingual chat, tool calling, 196K context, and reproducible local benchmarks.
- Cogni-Brain Qwen Fast: Fast local tool-agent brain for sandboxed DGX Spark workflows: Qwen 3.6-35B via Atlas with 218.85 tok/s, 100/100 Tool-Eval, and local NVFP4 acceleration for responsive Claude Code and agent loops.
- Cogni-Brain Nemotron: Original large local reasoning brain for sandboxed DGX Spark workflows: Nemotron-3-Super-120B via vLLM with stable 131K local context, ~24 tok/s, 93/100 Tool-Eval, hardened agentic stack, and reproducible benchmark methodology.
- SageGPT-7.5M: Small language model trained from scratch on ~140M pure Sanskrit tokens on NVIDIA DGX Spark. 6 Layer, 8 Attention Head, 256 embed, 1024 context.
- SageGPT-7M MLX: Small language model trained from scratch on ~57M Sanskrit tokens on Apple Silicon. 4 Layer, 8 Attention Head, 256 embed, 256 context.
- Cogni.chat: Local-first multimodal AI ecosystem for personal and household intelligence, designed to support memory, planning, wellbeing, learning, creative work, family coordination, and multimodal interaction across text, voice, images, and personal context.
- Fiduciary-Ops-Agent: Autonomous enterprise governance agent utilising a strict Check-then-Act protocol via Gemini 2.5 Flash Lite; enforces real-time fiduciary risk-alignment using tool-first orchestration.
"I have no special talent. I am only passionately curious" — Albert Einstein