Building
A personal initiative to build practical, secure, and sovereign tools, agents, and infrastructure setups within AI. Focus on utility, privacy, and speed.
Showcase Legend & Key
Currently Building
Local Email Archive Semantic Search
Migrating historical IMAP mail archives to Maildir and indexing with notmuch CLI search tool powered by local LLMs and embeddings.
CodeTour Generation MCP Server
Building an MCP server that parses git changesets and codebase structure to automatically write interactive CodeTour (.tour) walkthroughs.
Direct Filesystem Search Optimization Benchmark
Evaluating direct OS filesystem read/mmap and local ripgrep integration to bypass abstraction overhead and accelerate searches.
Projects
Your project?
Need to accelerate your AI delivery without leaking proprietary data to third-party clouds? Let's build your next milestone together.
AMD ROCm Stack Compilation & llama.cpp Benchmarking
Compiled ROCm 7.2.4 on AMD Ryzen AI APU (Radeon 880M iGPU) for llama.cpp HIP/Vulkan benchmarking. Analyzed memory bandwidth bottlenecks.
AI Developer Tool Security Audit
Comprehensive threat modeling and auditing of developer tools (Cursor, Cody, Aider). Exposed plaintext token storage and exfiltration risks.
AI FinOps & Token Economics Model
CapEx vs OpEx analysis of local eGPU/unified memory setups against serverless cloud API token usage, modeling agent-debate context inflation.
Local Git History AI Summarizer
Developed a local CLI tool that generates semantic, structured git commit summaries and release notes using a private 3B coding model.
Local Email Archive Semantic Search
Migrating historical IMAP mail archives to Maildir and indexing with notmuch CLI search tool powered by local LLMs and embeddings.
CodeTour Generation MCP Server
Building an MCP server that parses git changesets and codebase structure to automatically write interactive CodeTour (.tour) walkthroughs.
GPG-Encrypted AI Output Secure Protocol
Implementing a CEO-mandated secure output protocol (ADR 013) that automatically GPG-encrypts all agent-generated scripts and reports before storage.
Agent Command Audit Log Generator
A compliance-driven pre-commit agent hook that automatically records, formats, and registers all executed CLI commands in daily audit files.
CrewAI Python Multi-Agent Folder Classifier
A multi-agent file classification system utilizing CrewAI (Intake, Analysis, Decision, and Cleanup agents) to evaluate local Linux directories and suggest compliance actions.
C# AutoGen.Ollama Folder Classifier Port
Ported the CrewAI folder classifier to C# .NET 8.0 using Microsoft Agent Framework and AutoGen.Ollama to run local offline classifications.
DORA Delivery Metrics & Telemetry Calculator
A local agent utility and CLI tool that parses Azure DevOps pipelines (azdo-pipe) and GitHub Actions logs to calculate live DORA delivery metrics (Deployment Frequency, Lead Time, Change Failure Rate, MTTR) and business outcome telemetry.
Automated Agent Persona Avatar Generator
A local image generation pipeline utilizing stable diffusion models to create consistent, high-quality, and stylized avatar images for SRE agent personas.
eBPF Observability Agent Persona
Implementing virtual eBPF specialist personas (Thomas Graf, Liz Rice) capable of parsing network flows and system calls to diagnose performance anomalies.
Rust Evangelist Agent Persona (Ferris)
Creating a Rust evangelist persona in the local agent registry to guide developers on safety-by-design, memory safety, and clean Rust architecture patterns.
LowLevelTV & Dr. Jonas Birch Agent Personas
Deploying expert low-level C and Vulkan/C++ engine specialist personas into the multi-agent design arena to debate performance optimizations.
Direct Filesystem Search Optimization Benchmark
Evaluating direct OS filesystem read/mmap and local ripgrep integration to bypass abstraction overhead and accelerate searches.
ainytta.blog: Agent-Native Sovereign Publishing Platform
Sovereign, zero-framework Astro 7 static publishing platform engineered for human craftsmanship and autonomous AI agent consumption (dynamic /llms.txt discovery, raw Markdown endpoints, public executable Agent Skills, and North Star governance).