Logo
Neural Language Systems

LLM & Generative
AI Projects

Exploring the frontier of large language models — from RAG-based knowledge systems to parameter-efficient fine-tuning and collaborative multi-agent architectures.

3Projects
RAG · LoRA · AgentsTechniques
Ethiopian AIDomain
AI chat interface on dark screen, deep navy background, faint cyan glow from monitor, atmospheric low-light office environment
RAG System
🇪🇹
Active

Ethiopian AI Assistant

RAG-based assistant with persistent memory

A retrieval-augmented generation assistant built specifically for Ethiopian knowledge domains — history, culture, language, and geography. Uses vector search over curated Ethiopian documents with conversation memory.

  • Ethiopian cultural knowledge base
  • Persistent conversation memory
  • Gradio web interface
  • Amharic + English support
RAGGradioLangChainFAISSHuggingFace
Neural network training visualization, dark background with purple data streams, dim atmospheric lighting, deep shadow environment
Fine-tuning
🧬
Active

Fine-tuned Ethiopian Domain Model

LoRA fine-tuning of TinyLlama on Ethiopian culture

Applied parameter-efficient LoRA fine-tuning to TinyLlama-1.1B on a curated dataset of Ethiopian cultural texts, historical records, and regional knowledge. Achieves domain expertise without full retraining.

  • LoRA parameter-efficient training
  • Ethiopian cultural dataset
  • Domain-specific reasoning
  • Deployable on consumer hardware
LoRATinyLlamaPEFTHuggingFacePyTorch
Abstract network of connected nodes on dark background, deep shadows, faint green connection lines, dim atmospheric lighting
Multi-Agent
🤝
Active

Multi-Agent Research Assistant

Collaborative agent system

Three-agent pipeline where a Researcher agent gathers and summarizes sources, a Writer agent drafts content, and a Critic agent reviews and refines the output. Built with LangChain agent framework.

  • Researcher + Writer + Critic agents
  • Iterative refinement loop
  • Source citation tracking
  • Configurable agent personas
Multi-AgentLangChainPythonOpenAI API
🧠

Ethiopian-Focused AI Research

All three projects are designed with Ethiopian language, culture, and knowledge domains in mind — bridging the gap between global AI capabilities and local context.