Levi DeHaan

AI Projects

Explore local inference engines, AI agents, model-serving systems, and applied machine learning projects by Levi DeHaan.

  • AI Agent and IDE Training Initiative

    Strategic initiative to integrate AI-powered development tools into developer workflows, with focus on Windsurf IDE and AI agent training for enhanced productivity

  • AI Trends Knowledge Graph

    Transform customer support data into actionable insights with advanced AI-powered analytics. Turn your customer support tickets into a powerful knowledge graph that reveals hidden patterns, customer behaviors, and emerging trends.

  • AI Cluster — Local AI Appliance

    Go runtime and control plane for a local AI appliance: model loading, compatible inference APIs, scheduling, durable chat, agents, projects, and a Svelte UI — with Vault for identity/connectors and Helix for encrypted knowledge.

  • Almagest — Native Inference Engine

    C++ llama.cpp engine with durable native workflows, layered memory, MCP tools, and a live dashboard. Runs Qwen3.8-27B across two RTX 3060s with MTP speculation.

  • AI-Powered Code Intelligence Platform

    Comprehensive Git repository analyzer and test generator with OpenAI-compatible AI integration, visual analysis, and intelligent caching. Clones repositories, performs detailed code analysis, and generates intelligent insights and tests with a visual interface.

  • Petco Fine-Tune Workflow

    Comprehensive pipeline for creating, testing, and deploying custom AI models using LLaMA Factory, PEFT, gRPO, and dPO with distributed computing capabilities

  • FluentD LLM Server - Intelligent Log Analysis System

    End-to-end log analysis solution combining FluentD log aggregation with Large Language Models (LLMs) for real-time security threat detection. Features custom-trained binary classification models using Microsoft Phi-2 for automated incident response and system anomaly detection.

  • LlamaLoadBalancer - AI Model Load Balancer

    FastAPI-based intelligent load balancer for multiple Llama language models with Redis queues, PostgreSQL persistence, Prometheus metrics, and specialized model routing for finance, programming, log analysis, and creative tasks.

  • LocalAI Test Case Generator

    Sophisticated web application that streamlines software testing by automatically generating comprehensive test cases using LocalAI's language models, featuring AI-powered test generation, structured JSON output, and a modern React interface.

  • Petco PetChat - Enterprise AI Chat Interface Suite

    Enterprise-grade AI chat platform for Petco employees featuring secure Okta SSO, comprehensive RBAC, LocalAI multi-model integration, persistent chat history, and a Carbon Design System admin console built on Next.js 14.

  • Petco SimpleAI (S3 Model Loading)

    Heavily customized LocalAI fork with asynchronous S3 model loading, real-time status tracking, two-level queue architecture, health monitoring, CUDA 12.0 optimization, and advanced build system for GPU-accelerated AI model serving.

  • Simple AI Router (SAR)

    High-performance gRPC router for AI inference with intelligent load balancing, traffic filtering, and multi-cluster support

  • SimpleAI Nutanix GPU ISO - Custom Ubuntu ISO with Petco SimpleAI

    Custom Ubuntu ISO bundling Petco SimpleAI inference server with systemd service, automated build process using Vagrant and Makefile for rapid deployment across Nutanix GPU clusters with NVIDIA GPU support and cloud-init integration.

  • Stock Dip Recovery Analyzer - ML-Powered Trading Analysis Platform

    Comprehensive full-stack application for analyzing historical stock performance to understand dip recovery patterns using machine learning, real-time market data, advanced technical analysis, and predictive modeling with 95%+ data coverage and 300+ epoch training.

  • Voice Master - Advanced AI Voice Assistant

    Sophisticated multi-modal voice assistant with Whisper.cpp, Piper TTS, and advanced computer command processing. Features multi-model AI integration, project management, specialized agents, and high-performance audio processing with CUDA acceleration.

  • Fetch Watch AI - Luxury Watch Intelligence Platform

    My son Gideon (https://gideondehaan.dev/) and I built this together, it's a Complete watch sourcing and pricing automation platform with AI-powered scraping, real-time analytics, messaging parser, self-healing scraper system, and comprehensive dashboards built with React, Express, PostgreSQL, Redis, and OpenSearch.

  • YouTube Financial Video Analyzer

    Comprehensive AI-powered financial analysis platform for YouTube content with automated reporting and market intelligence