Jake Goldsborough

Software Engineer | DevOps & Automation

email: jobs@jakegoldsborough.com

website: https://jakegoldsborough.com

github: https://github.com/ducks

Versatile Software Engineer with over a decade of experience in software development, DevOps, and automation. Proven ability to design, develop, and optimize scalable systems, reducing operational overhead and improving performance. Passionate about problem-solving, automation, and infrastructure as code. Experienced in working remotely and collaborating with cross-functional teams to deliver efficient, maintainable solutions.

Experience

Discourse

Software Engineer (Infrastructure), 2025-current

  • Eliminated 16x storage inflation on backups via upload deduplication. Shipped a hardlink-based deduplication pass in Discourse's backup creator that groups uploads by content hash and stores duplicate content once. Reduced one customer site from 432 GB to 26 GB of unique content. Followed up with an ext4-hardlink-limit fallback after finding a single reaction GIF duplicated 246,173 times, turning what would have been 246k redundant S3 downloads into 5.
  • Designed pre-provisioning pool for instant trial signups. Architected and shipped a pool of pre-built Discourse sites claimed at signup time, using FOR UPDATE SKIP LOCKED for safe concurrent claiming with graceful fallback to live provisioning. Target signup time of 90s reduced to under 15s. Includes a subdomain-name generator producing 2,805 readable combinations from curated word lists.
  • Optimized Ruby/SQL hot paths in core and internal services. Rewrote Discourse's daily category-stats job from per-row Ruby iteration (8 queries per category) to batched SQL aggregation with COUNT(*) FILTER (WHERE ...), achieving an 8.6x speedup at production scale (2,000 categories / 1.6M posts). Separately refactored an internal hosted-sites API endpoint from 27-second timeouts to under 10 seconds by eliminating N+1 Redis lookups and adding a request-scoped tier cache.
  • Parallelized CI and restructured build pipelines. Cut a Ruby/Rails service's CI runtime from 10 minutes to 3-4 by parallelizing RSpec across 4 GitHub Actions runners, surfacing and fixing several pre-existing test-isolation bugs in the process. Restructured the build Dockerfile into multi-stage layers so lockfile changes no longer bust source-code caches, cutting code-only deploys from ~10 minutes to ~6.
  • Contributed S3 IAM role assumption to Discourse core. Multi-quarter upstream contribution letting Discourse installations use IAM role assumption for S3 instead of long-lived access keys. Worked around a Ruby AWS SDK credential-chain precedence bug, refactored a singleton that was leaking credentials across sites in multisite mode, and collaborated with maintainers through review iteration on a safer simplified release.

Flashpoint

Software Engineer II (Mid-Level), 2022-2024

  • Reduced infrastructure setup time from days to hours. Led Terraform and Kubernetes pipeline implementation. Designed reusable Terraform modules and automated Kubernetes resource creation using Node.js. Initiated Helm chart development for better deployment management.
  • Refactored a legacy Java-based API into a modern Node.js service. Reduced technical debt and simplified maintenance for future feature requests.
  • Mentored junior engineers. Contributed to technical interviews, fostering a culture of continuous learning and collaboration.

Infinity Interactive

Software Engineer, 2010-2020

Contracted full-stack work for Fortune 100/500 clients (Shutterstock, Allegra, Flashpoint). Node.js, GraphQL, React, Vue.js. Improved a high-traffic app's response times by 75% via a Redis caching strategy.

Projects

claux

https://crates.io/crates/claux

Terminal AI coding assistant in Rust. Streaming chat with tool execution (Read, Write, Edit, Grep, Bash, WebFetch, Agent), interactive permissions, session persistence, auto-compaction, sub-agents, multi-provider routing (Anthropic, OpenAI, Ollama), and a ratatui TUI.

llm-mux

https://crates.io/crates/llm-mux

Multi-backend LLM router and orchestrator. Routes prompts across CLI and HTTP backends with parallel execution and multi-step workflows.

finna

https://crates.io/crates/finna

Multi-model debate-to-implement pipeline. Three models independently propose, Claude synthesizes consensus, breaks it into dependency-ordered specs, applies edits.

Skills

Cloud & Infrastructure: AWS, Terraform, Kubernetes, Google Cloud

Databases & Caching: Postgres, MySQL, Redis, GraphQL

DevOps & CI/CD: Docker, Git, NixOS

Frontend: React, Vue.js, AngularJS

Languages: English (Native), Swedish (Learning)

Operating Systems: Linux/Unix (Debian, Ubuntu, NixOS, macOS), BSD

Programming: Ruby, Node.js, TypeScript, Python, Rust