Lab Projects — PYXON Labs

EdTechEdge AI

Pyxon-Eye LMS Grading for Moodle

Integrate Pyxon-Eye AI into LMS platforms like Moodle for question-by-question grading and analysis—running on limited GPU machines with Pyxon SLLM.

Arabic NLPBenchmarks

Arabic Reasoning Search

Embedding retrieval plus LLM pairwise re-ranking with Arabic explanations for high-precision semantic sentence search.

Edge AIBenchmarks

On-Device RAG Benchmark

Edge RAG that lifts weak local LLMs to strong device-manual QA accuracy without cloud round-trips.

Edge AIArabic NLP

Pyxon SLLM v0

A compact ~1.6 GB Arabic instruction model packaged for offline and edge use (including Ollama).

Safety & DetectionArabic NLP

Arabic AI Text Detector

High-accuracy detection of AI-generated vs human Arabic text for authenticity and trust workflows.

BenchmarksArabic NLP

Arabic Poetry Benchmark

Meter and rhyme evaluation for specialized Arabic generation—exposing gaps in general LLMs.

GovernanceSafety & Detection

Country AI Governance Scoring

Score LLMs and RAG systems against country AI rules (e.g. SDAIA, UAE, EU) with developer-friendly packages.

Edge AIBenchmarks

GPU Lab

Map GPUs to LLM inference and fine-tuning capacity so teams pick hardware with evidence, not guesswork.

Safety & DetectionAgents

LLM Security Red Team

Prompt injection, token-drain, and adversarial testing to harden Arabic and enterprise LLM deployments.

Arabic NLPAgents

Modular RAG System

Composable RAG building blocks—Search, Fusion, Memory, Routing, Predict, Task Adapter—for Arabic and enterprise stacks.

Edge AIArabic NLP

Edge LLM Distillation

Compress Arabic-capable models for mobile and CPU inference without abandoning regional quality.

AgentsArabic NLP

Arabic Agentic Workflows

Approval-aware agents grounded on PYXON embeddings for multi-step Arabic enterprise missions.

Governance

ISO 42001 AIMS Readiness

Adaptive interview that scores AIMS maturity and produces clause gaps plus a 90-day action plan.