Skip to content

Joseph Ellis Stewart

Principal Systems & Software Architect

Building lean systems, embedded runtimes, array-oriented synthesis engines, and telecommunications infrastructure.

About

I am a Principal Systems and Software Architect specializing in deterministic, low-latency computing environments, bare-metal embedded runtimes, and high-throughput telecommunications pipelines. My work bridges low-level hardware constraints with high-order algorithmic rigor—from millimeter-wave satellite payloads down to bare-metal audio synthesis primitives.

Over the past decade and a half, I have spearheaded the design and delivery of mission-critical systems across aerospace, communications, and digital signal processing. At Meta R&D, I guided the flight software architecture and ground verification pipelines for mmWave satellite constellations, ensuring sub-millisecond telemetry loops and deterministic link performance under extreme physical and channel constraints.

Outside of commercial telecommunications, I engineer deterministic audio synthesis engines in pure C99, explore array programming paradigms for real-time DSP, and craft minimalist toolchains designed to run with zero unnecessary abstraction. I believe in software that respects memory hierarchies, provides predictable latency bounds, and delivers uncompromising reliability.

Experience

  1. 2021 — PRESENT

    Architected core flight software and ground-station verification suites for next-generation mmWave satellite payloads. Spearheaded deterministic telemetry acquisition protocols, beam-steering state machines, and real-time radio calibration workflows. Collaborated across RF hardware, FPGA, and embedded firmware engineering teams to guarantee sub-millisecond command-and-control dispatch and zero-drop link telemetry.

    • C++20
    • C99
    • Linux RT (PREEMPT_RT)
    • mmWave Telemetry
    • ZeroMQ
    • SDR / GNU Radio
    • eBPF
    • Python
  2. 2017 — 2021

    Lead Systems & Infrastructure Architect · Telecommunications & Embedded Systems

    Engineered resilient edge computing runtimes and packet processing pipelines for low-latency carrier networking. Directed migration toward custom lock-free memory rings and SIMD-accelerated packet inspection, driving a 4x reduction in 99th-percentile transit latency. Standardized automated HIL (Hardware-in-the-Loop) test harnesses across multi-site lab environments.

    • C
    • Rust
    • DPDK
    • POSIX RT
    • gRPC
    • Automated HIL
    • Kernel Networking
  3. 2013 — 2017

    Senior Embedded Software Engineer · Digital Signal Processing & Real-Time Runtimes

    Designed embedded signal processing kernels and DMA-driven audio and RF ingestion pipelines. Authored hard real-time interrupt handlers, circular buffer topologies, and cross-platform synchronization primitives deployed to ARM Cortex-M/A microcontrollers and DSP coprocessors.

    • ARM Cortex-M4/M7
    • FreeRTOS
    • Fixed-Point DSP
    • Bare Metal
    • I2S / SPI / UART
    • C99

Projects

  1. A zero-dependency, pure C99 polyphonic wavetable synthesis engine built for hard real-time audio generation. Features SIMD-vectorized band-limited wavetable interpolation, lock-free ringbuffer command queuing, multi-channel voice stealing, and a decoupled miniaudio backend. Compiles cleanly to native binaries and WebAssembly (WASM/WebAudio) with UDP command dispatch for headless live control.

    • Pure C99
    • miniaudio
    • WebAssembly
    • WebAudio
    • UDP Dispatch
    • SIMD DSP
    • Lock-Free Rings
  2. High-performance telemetry multiplexer and frame parser implementing custom binary framing for high-bandwidth satellite downlinks. Leverages memory-mapped rings, zero-copy socket buffers, and lock-free thread topologies to process multi-gigabit streams with deterministic jitter under 15 microseconds.

    • C++20
    • Kernel Bypass
    • Memory-Mapped I/O
    • Zero-Copy
    • POSIX RT
  3. A concise, array-oriented numerical engine optimized for multi-channel spectral decomposition and polyphase filterbank synthesis. Developed to investigate cache locality patterns and auto-vectorization efficacy across ARM NEON and x86 AVX2 microarchitectures without external BLAS dependencies.

    • C99
    • AVX2
    • ARM NEON
    • Fourier Transforms
    • Cache Locality