Kyle Bradshaw

Full-stack engineer — React, Go and Python microservices, and LLM/RAG integration. Four years of experience, the last stretch spent consulting and building production systems independently: designing the APIs, shipping the frontends, and running the whole stack on Kubernetes. Everything below is deployed and instrumented, not a demo.

Every service in this portfolio ships Prometheus metrics to a live Grafana dashboard.

Featured Project

GalaxyVoyagers.com

Deployed collaborative sci-fi worldbuilding platform with a Go GraphQL gateway and AI-assisted creation tools

Built with Next.js, Apollo Client, Go, gqlgen, gRPC, PostgreSQL, MongoDB, Redis, RabbitMQ, and AI generation. View the architecture walkthrough for the full system design.

Backend & Data Engineering

Go Ecommerce Platform

Microservices ecommerce platform built with Go, PostgreSQL, Redis, and RabbitMQ

Microservices architecture with JWT authentication, product catalog, cart, orders, and asynchronous worker pools — deployed on Kubernetes.

Full-Stack Java

Task Management System built with Spring Boot, GraphQL, and Kubernetes

Microservices architecture with PostgreSQL, MongoDB, Redis, RabbitMQ, Google OAuth, and CI/CD automation.

Database Engineering

Production PostgreSQL — pooling, replication, optimization, partitioning, migration safety, and reliability

Real benchmarks against PostgreSQL 16, transaction-mode PgBouncer pooling, an async streaming read replica with a separate reporting pool, range partitioning with materialized views, a custom AST-based migration linter, and verified point-in-time recovery.

Asynchronous Systems

Go ecommerce messaging with Kafka event streams, RabbitMQ sagas, DLQs, replay, and production observability

Checkout saga command/reply queues, bounded retries, publisher confirms, reconnect-aware RabbitMQ publishing, Kafka-backed order events, CQRS projection, streaming analytics, DLQ envelopes, and traceable recovery paths.

AI Systems

Document Q&A Assistant

A RAG document assistant plus a Kafka-scale sensitive-data classifier — retrieval-augmented generation and applied LLM classification

A full-stack retrieval-augmented generation system (FastAPI, Qdrant, Ollama) for PDF Q&A, plus a DSPM classifier that detects sensitive data at Kafka scale with a tiered regex → NER → LLM pipeline. Explore both from the AI page.

Platform & Operations

Observability

Production-journey instrumentation — Prometheus metrics, Loki logs, Jaeger traces, and live alerting

Three-pillar stack with deploy annotations, Kubernetes event exporter, gRPC client interceptors, saga-stalled alerts, and Kafka-header trace propagation across the async boundary.

CI/CD Pipeline

Unified GitHub Actions workflow with a live QA environment at qa.kylebradshaw.dev for pre-prod review

A single workflow handles quality checks, image builds, and deployments for three service stacks — designed for a solo developer with automated spec-to-production delivery. See what's currently staged for production review on the CI/CD page.

Infrastructure & Deployment

Production Kubernetes on a home server, AWS-ready with Terraform and EKS

Two deployment architectures for the same services — a cost-effective Minikube cluster with Cloudflare Tunnel serving production today, and a one-command AWS deployment with EKS, RDS, ElastiCache, and Amazon MQ.

Security

Defense-in-depth across the stack — application, CI/CD, Kubernetes, and the hardened Linux host that runs it all

Six CI security gates, JWT + httpOnly cookies, pod security contexts, Sealed Secrets for GitOps-friendly secret management, UFW default-deny firewall, Tailscale-only SSH, auditd, sysctl hardening, and a lynis baseline score of 77.