Gemma 4 31B Abliterated — quality-preserving guardrail removal for Google's most capable open model. Apache 2.0. Runs on Apple Silicon via MLX.

Personalize Gemma 4 and make it real-time on Apple Silicon. Fine-tune on your conversations, serve at 111 tok/s. No cloud required.

Domain-specialized Gemma 2 27B + Gemma 4 31B for SEC filings — fine-tuned on TPU v6e-8 with PyTorch/XLA FSDPv2, plus a Vertex AI Vector Search RAG demo (69 tickers × 381 filings). Same LoRA recipe, +3.5% / +5.8% BERTScore F1.

Offline subtitle translator with 2026 LLMs (Gemma 4, Qwen 3.5, Hunyuan-MT, Llama 4 Scout) via Ollama. Web UI + CLI, 17 languages, video subtitle extraction, persistent Translation Memory, LLM-as-judge quality estimation, genre-aware prompts. No API keys, no cloud.

Benchmark Gemma 4 E2B on Apple Silicon: MLX (mlx-lm) vs GGUF (llama-server), with TTFT, tokens/sec, and memory.

Drop-in KV cache compression for MLX on Apple Silicon. Brings PolarQuant (Google, ICLR 2026) to mlx-lm with first-class Gemma 4 support: MatFormer, dual head_dim, hybrid sliding/global attention, cross-layer KV sharing. 3-bit → 4.8× smaller cache, 0.995 logit cosine @ 4-bit.

Production-ready local AI agent stack for NVIDIA DGX Spark / ASUS GB10. Gemma 4 31B NVFP4 + bge-m3 embeddings + OpenClaw gateway, one docker compose up.

M.A.R.A. is a Multi-Agent Research Analyst. A self-correcting multi-agent research system built with LangGraph, Gemma 4, and Streamlit.

Hands-on benchmark of Google Gemma 4 in LM Studio, with prompts, Gradio evaluator, saved outputs, and a published companion report.

Gemma-4-It fine-tuned on PubMedQA using SFT & RLVR

Nairobi-first edge AI simulation for urban flood resilience. Orchestrating multimodal agents (Gemini 2.5/Gemma 4) to navigate the Mbagathi Basin during infrastructure failure. Offline-first, spatial-aware, and mission-critical.

Quantum-assisted verification of LLM reasoning. Translates each step to first-order CNF, grounds against a finite universe, then checks consistency with Grover's search on PennyLane simulator or IBM Heron r2 hardware.

A local-first natural-language Q&A engine for EU directives using on-device LLMs, hybrid search (Qdrant), and citation graphs.