Clarence Bowen

Systems & Data Engineer • FPGA (SystemVerilog) • Low-Latency Systems

I build data systems that move and transform data efficiently, from production pipelines and APIs to low-level ingestion and memory paths. Production experience with SQL transformations, ETL/data pipelines, and API-driven systems; focused on reliability, performance, and system correctness. Also exploring low-level systems and FPGA-based data paths.

Open to consulting, technical architecture reviews, and focused systems/data engineering engagements.

Work Inquiry LinkedIn Blog

~ / explore_by_domain

Applied Systems Architecture

Data pipelines and API-driven systems.

FPGA / Low-Latency Systems

RTL, AXI streaming, and lock-free memory pipelines.

Research / Writing

Deep-dives into systems thinking, data semantics, and design constraints.

> Core Architecture & Systems

Data systems, APIs, and SQL-driven DAG workflows.

OpenCurb

Spatio-Temporal Database / Geospatial API

A data system that models curb regulations as a queryable API. Given inputs such as location, time range, vehicle type, and other parameters, it returns what actions are allowed.

(lat, lon, vehicle_type, ...) -> allowed curb actions

Encodes complex municipal rules into a unified query model for routing and compliance.

View Live System ↗

> High-Throughput Pipeline

Technical case study: Reducing OS overhead to improve ingestion throughput

Level 1: The Hardware Plane

FPGA / RTL

Packet Formatter + AXI DMA Path

A SystemVerilog cut-through formatter on a Zynq-7000 that formats incoming AXI4-Stream packet data into fixed-width 32-byte records. Designed to feed AXI DMA into memory, producing structured records for downstream data processing in a cache-aligned SPSC ring buffer.

Level 2: The Memory Plane

C

Cache-Aligned SPSC Ring Buffer

A pure C lock-free SPSC ring buffer designed to consume fixed-width records from a memory-mapped DMA buffer (planned integration). Prevents MESI-induced false sharing by padding atomic read/write indices to 64-byte L1 cache-line boundaries.

Throughput: ~18–28M msgs/sec (~3–5× faster than mutex queue)

View Benchmark Repo ↗

Writing & Research

FunDataMentals.com

Blog on first-principles and bridge-principles systems thinking, connecting human behavior, software, hardware, and meaning. I write about uncertainty, data semantics, systems design, and the hardware/software boundary.