Minimal CPU-only Ollama Docker Image

🦙 chat-o-llama: A lightweight, modern web interface for AI conversations with support for both Ollama and llama.cpp backends. Features persistent conversation management, real-time backend switching, intelligent context compression, and a clean responsive UI.

An LLM-based content moderator. Firefox extension to block webpages unrelated to work, based on page title and URL. Local LLMs with Ollama and Langchain to ensure your browsing history never leaves your device, for complete privacy. Google Gemini also supported.

A high-performance Python library for extracting structured content from PDF documents with layout-aware text extraction. pdf_to_json preserves document structure including headings (H1-H6) and body text, outputting clean JSON format.

Image Classification with On-Device Inference, built with Flutter, AI model runs on mobile cpu

Ternsig Virtual Mainframe Runtime (TVMR) — extensible VM with 10 standard extensions (121 instructions), Signal ISA, mastery learning, hot-reload firmware, and thermogram persistence.

Cloud transcription stores your audio. Local alternatives need a GPU. Whiscribe runs on CPU, in your browser, with one Python script.

Face locking system built on ArcFace (ONNX) and 5-point alignment that recognizes a selected identity, locks onto it, tracks facial actions, and records behavior over time.

Hybrid RAG on CPU-only SQLite: FTS5 + embeddings + RRF fusion. 235 knowledge sources.

CPU-friendly experience-based reasoning framework combining meta-learning (MAML), state space models (SSM), and memory buffers for fast few-shot adaptation. Pure NumPy implementation for edge devices and low-compute environments.

Real-time sign language detection & translation using MediaPipe, LSTM, and Gemini 2.5 Flash — with WebSocket streaming, TTS audio output, and a Next.js frontend. CPU-only, low-spec friendly.

A new one shot face swap approach for image and video domains - version tailored to work on CPU

Pre-built Llama-CPP Wheel for HF Spaces (Python 3.13)

Probabilistic Signed Distance Fusion with View Planning on CPU

One-line TTS for Python. Real speech on any CPU. No GPU, no cloud, no API keys.

CPU-optimized RAG pipeline reducing latency 2.7× (247ms → 92ms). Implements caching, filtering, quantization for production. Complete with FastAPI, Docker, benchmarks, investor materials. The engineering showcase that sells itself.

Train AI together — any device, any hardware. Federated fine-tuning with LISA layer selection.

CPU-only RAG stack: PDFs→Docling→Ollama→pgvector. Windows/macOS/Linux. Docker Compose. Graph-aware code search + scanned PDF OCR.

A lightweight reproduction and analysis inspired by recent work on presentation-aware deepfake / spoofing detection, with a focus on codec-induced presentation mismatch (AMR) under CPU-only constraints.