SYSTEMS ARCHITECTURE · DEVELOPER TOOLS · APPLIED AI

Developer Tooling for AI Agents, Low-Latency Protocols & Hardware Systems.

Desktop developer tools for AI coding agents, sub-millisecond terminal stream parsers, native Bluetooth hardware drivers, and local inference setups. Focused on observable state, fast native code, and systems that run reliably without cloud lock-in.

Explore Engineering Projects arrow_downward mail [email protected] View Background on LinkedIn arrow_forward

Desktop AI IDE

Copernicus

Agent workspace & inspection UI

Terminal Filter

< 1.0 ms

ANSI / VT100 stream parsing

Bluetooth Driver

Swift 6

CoreBluetooth thermal printer driver

Local Event Store

SQLite WAL

Crash-safe hardware event queue

Technical Capabilities

Core Competencies & Toolkit

Engineering domain expertise built on deep systems fundamentals and modern AI tooling.

smart_toy

AI & Agent Tooling

Developer environments for coding agents: context management, custom Model Context Protocol (MCP) servers, tool-calling shims, and local models via MLX, CoreML, and Ollama.

MCP MLX / CoreML Ollama Python

terminal

Protocols & Terminal

Low-latency terminal stream filtering, POSIX pseudo-terminal (PTY) handling, VT100/ANSI escape sequence parsing, circular buffer management, and real-time screen diffing.

C / POSIX VT100 Swift PTY

devices

Hardware & Drivers

Reverse-engineering Bluetooth Low Energy (BLE) peripherals, writing native Swift CoreBluetooth drivers, Floyd-Steinberg image dithering, and SPI/I2C sensor integration.

Swift 6 CoreBluetooth PN532 SPI/I2C

sync_alt

Enterprise Integrations

Connecting disparate business systems: healthcare and supply chain messaging (HL7, ANSI X12, EDI), REST APIs, ETL pipelines, and concurrent SQLite WAL event logging.

REST APIs SQLite WAL EDI / HL7 ETL

Portfolio Showcase

Featured Engineering Projects

Open-source repositories and standalone system implementations highlighting clean architecture, protocol design, and performance.

Flagship Platform · Open Source Public Repository

Copernicus IDE — Desktop Workspace for AI Coding Agents

An Electron and React desktop application for running and observing AI coding agents locally. Gives developers a clear UI to inspect tool execution steps, view file diffs, and manage context windows instead of watching an unformatted terminal stream.

Step Inspection & Replay Logs every agent action, tool input, and output so you can inspect past steps and debug agent missteps.

Local Python IPC Electron UI talks to a Python runtime over local UNIX domain sockets using structured JSON messages.

Local & Cloud LLMs Supports local inference with Ollama alongside cloud APIs, allowing offline agent execution on Apple Silicon.

Electron React 18 TypeScript Python IPC Ollama

View on GitHub open_in_new

Systems & Protocol Engine Public Repository

Claude Code Barber & yasr — Fast Terminal Stream Filter

A lightweight C and Swift utility that intercepts raw pseudo-terminal (PTY) output, cleans up ANSI escape codes and progress bars, and extracts stable screen text. Ported from the open-source yasr screen reader to turn noisy interactive terminal streams into clean text for AI agents.

Sub-Millisecond Latency Uses circular memory buffers to process megabytes of terminal output in microseconds without slowing down the host process.

VT100 State Machine Tracks terminal cursor movements, line clears, and color codes to reconstruct the visible screen state.

Native POSIX PTY Spawns and monitors child processes through standard UNIX pseudo-terminals with zero external dependencies.

C / POSIX Swift VT100 Emulation Zero-Copy Buffer

View on GitHub open_in_new

Hardware & BLE Driver Public Repository

CatPrinter Swift — Native macOS/iOS Thermal Printer Driver

A standalone Swift 6 command-line tool and macOS daemon that prints text, barcodes, and images to portable Bluetooth thermal printers. Built by reverse-engineering BLE packet traces, with zero external dependencies.

Pure Swift 6 & CoreBluetooth Uses modern Swift concurrency (async/await) and Apple's native CoreBluetooth APIs directly—no third-party dependencies.

Bitmap Image Dithering Converts grayscale photos and graphics into sharp 1-bit monochrome bitmaps using Floyd-Steinberg error diffusion for 203 DPI print heads.

Binary Protocol Framing Packages image lines into binary payloads with CRC-8 checksums and hardware control opcodes (0x51 0x78).

Swift 6 CoreBluetooth BLE GATT Dithering Python Daemon

View on GitHub open_in_new

Hardware & Embedded Systems Architecture Spec

NFC Edge Inventory Pipeline

Physical asset and badge tracking station powered by a PN532 NFC module connected over SPI/I2C. Reads tags in under 50ms, handles audio feedback, and writes scans locally to a crash-resilient SQLite database.

• Fast anti-collision handling for overlapping tags

• Offline-first local queue that syncs when network is available

• SQLite Write-Ahead Logging (WAL) for safe, concurrent writes

PN532 NFC SQLite WAL Python 3.12 SPI / I2C

Local Inference & Infrastructure Infrastructure Spec

Local AI Fleet Orchestration

A multi-node local setup running open-source models across Apple Silicon hardware and workstation GPUs. Uses custom Python middleware to route tasks across MLX, CoreML, and llama.cpp/Ollama runtimes based on context size and memory constraints.

• MLX & CoreML acceleration leveraging Apple Silicon unified memory

• Lightweight routing shim for agent tool execution across local models

• Air-gapped model execution with zero cloud telemetry or data egress

Apple Silicon MLX CoreML llama.cpp Ollama Tool Shims

Systems Philosophy

Engineering Principles

The architectural standards applied across every project, tool, and system integration.

01 / OBSERVABILITY

Debuggable State & Clear Logs

Whether coordinating AI agents or moving data between enterprise databases, every operation should be recorded in a clear log. If an unexpected failure occurs, you should be able to inspect the exact inputs and replay the event.

02 / EFFICIENCY

Lean, Native Code

Avoid heavy frameworks when native OS primitives do the job better. Using direct POSIX calls, standard UNIX sockets, and memory buffers produces software that is faster, uses fewer resources, and has fewer breaking dependencies.

03 / RELIABILITY

Works Offline & Stays Up

Systems shouldn't fail simply because an internet connection dropped. Designing with local SQLite stores, queued retries, and on-premise inference ensures tools remain functional even when upstream services are down.

04 / DOCUMENTATION AS CODE

Docs as a Core Deliverable

Code and user documentation are treated with the exact same rigor as production code. Clean architectural specs, internal API references, and step-by-step user guides eliminate tribal knowledge and keep codebases accessible.

05 / PLANNED EXECUTION

Spec & Backlog-Driven Outcomes

Consistent engineering outcomes are built on disciplined upfront planning. Breaking initiatives down into PRDs, user stories, prioritized backlogs, and atomic tasks ensures predictability before writing a single line of code.

GET IN TOUCH

Let's Talk Systems, Tooling & Engineering

Looking for someone who builds pragmatic developer tools, low-level protocols, or enterprise integrations? Send an email or connect on LinkedIn to discuss roles, projects, or background.

mail [email protected] person Connect on LinkedIn code GitHub (@thrtysxty)

Focus: Developer Tooling · Hardware Protocols · Systems Integration

Professional Background: Available on LinkedIn