Max Baines - Portfolio

Virtual Modular Synthesizer

A real-time audio synthesis application built from scratch in Zig, allowing users to create and connect modular synthesizer components in a visual interface.

Overview

This project is a fully functional modular synthesizer with a node-based visual interface. Users can patch together oscillators, filters, effects, and other DSP modules to create custom synthesizers and generative music.

Architecture

The project is structured in three layers:

Core Library (scrublib)

Cross-platform foundation handling window management, rendering, and OS interaction:

• GLFW/GLAD for window creation and OpenGL context
• Custom immediate-mode renderer for shapes and text
• FreeType for font rendering
• RtAudio for cross-platform audio I/O
• RtMidi for hardware MIDI device support
• Input handling system

Audio Graph Engine

The heart of the synthesizer - a high-performance, headless audio processing system:

• Signal flow graph for node connections
• Polymorphic node system using a generic interface
• Type-safe ports for inputs/outputs with multiple signal types (audio, control, gates, integers)
• Memory pool architecture for zero-copy data sharing between nodes
• Cache-efficient buffer management for real-time performance
• Topological sorting for correct processing order
• Serialisation into JSON for saving/loading as well as copy/paste

GUI Layer

Visual interface for patching and control:

• Node-based visual programming interface
• Design allowing GUI to be decoupled from audio engine
• File browser for handling save files, samples, subpatches

Repository: Private repository - happy to provide access for review

Cool Beans Point Of Sale

A custom point-of-sale system built for a small coffee shop, demonstrating practical problem-solving with a simple, purpose-built solution.

Key Features:

• Product catalog with modifiers (sizes, milk alternatives, etc.)
• Printer support:
  • Customer receipts
  • Barista dockets for coffee orders
  • Kitchen dockets for food preparation
• SQLite database for product catalog and order history
• Simple, touch-friendly interface for fast service

Technical Implementation

• Platform: Godot Engine (GDScript)
• Database: SQLite with tables for products, product groups, modifiers, and order history
• Printing: Direct integration with Linux lp command for printing to network thermal printers
• Architecture: Separate docket printing allows kitchen and barista to work independently

Production Use

In production for 2+ years, processing daily orders with zero downtime or maintenance issues.

Repository: https://github.com/scrubmunch/CoolPOS

Adventures into Pen Plotters

Building a pen plotter machine and developing software to generate plotter art, exploring the intersection of computers and art.

What is a Pen Plotter?

A pen plotter is a computer-controlled drawing machine that moves a pen across paper using precise stepper motors. Unlike digital printers, plotters create artwork through actual pen strokes, giving them a unique aesthetic quality.

Hardware

Custom-built plotter:

• Aluminum extrusion frame with belts and stepper motors for X/Y movement
• Servo for pen lift mechanism
• Arduino running GRBL firmware that accepts G-Code commands
• Custom 3D printed parts

Software

Python-based generative art toolkit for creating plotter artwork:

Core Library Features:

• Primitive drawing functions (lines, curves, splines, shapes)
• Shape filling algorithms (hatching, stippling, concentric fills)
• SVG import and processing
• G-code generation and image preview system
• Path optimization - a unique variation on the travelling salesman problem

Generative Experiments

This core library enabled exploration of various areas of programming:

• Constraint solving algorithms
• Graph theory and path planning
• Fractals and recursive subdivision patterns
• Geometric tessellations and tiling patterns

Technical Stack

• Language: Python
• Key Libraries: Pyglet for graphics, Shapely for geometric operations
• Algorithms: Graph theory, computational geometry, TSP variants

Scribe - AI Medical Report Assistant

A full-stack web application that transforms audio recordings of patient consultations into medical reports using AI. Built as a freelance project for a client wanting to demo the idea to potential business partners. This was my first real dive into modern web technologies but I found I was able to pick up the skills quickly.

The Problem

Healthcare professionals spend significant time on documentation, manually transcribing recordings and formatting them into standardized reports.

The Solution

Scribe automates the medical documentation workflow by:

1. Converting audio to text using AssemblyAI's transcription service
2. Intelligently extracting and structuring information using OpenAI's GPT-4 to match customizable templates
3. Providing a polished editor for healthcare professionals to review and refine reports

Key Features

• Audio-to-Report Pipeline: Upload consultation recordings and receive AI-generated reports with streaming responses
• Custom Template System: Create and manage reusable report templates with placeholder fields that AI automatically populates
• Rich Text Editor: Built-in WYSIWYG editor for editing.
• User Authentication & Subscriptions: Multi-user management with Stripe-powered subscription service

Technical Stack

Frontend:

• Next.js 13
• React 18 with Server Components
• Tailwind CSS

Backend:

• Next.js API Routes (serverless)
• Supabase (PostgreSQL) for database and authentication
• AssemblyAI for audio transcription
• OpenAI GPT-4 for report generation with streaming responses
• Stripe for payment processing and subscription management