Gerard Caulfield

Senior Software Engineer / SRE · AI systems, real-time products, infrastructure

[email protected] · gerard.caulfield.me/hello · github.com/gerrywastaken

Profile

Software engineer with 15+ years across product engineering, SRE, cloud infrastructure, and team leadership. Currently building Sandegg, a real-time multilingual voice-calling product, alongside local-first AI assistant work. Strong bias toward production systems: latency, observability, deployment reliability, failure handling, and practical automation.

Selected Work

Sandegg · Founder / Solo Engineer August 2024 - Now

Building a real-time multilingual voice-calling product in private preview: product, voice pipeline, latency work, deployment path, and early user testing.
Validated end-to-end translated calls with Cantonese and English speakers, including live family conversations where neither side shared a language.
Concurrent R&D: local-first personal assistant with persistent memory and self-extending tools; mobile app iteration; geospatial visualization tooling.

Local-first Personal Assistant Private R&D

Desktop voice assistant with wake/command handling, cross-session memory, self-generated tools, local document/web search, clipboard access, and desktop context experiments.
Focuses on task execution, contextual recall, privacy, and low latency rather than simple chatbot interaction.

Experience

Senior Site Reliability Engineer · Workday August 2021 - 2024

Supported global applications across data centers, GCP, and AWS Kubernetes clusters.
Improved CI/CD systems, monitoring, alerting, and operational tooling used across teams.
Built automatic PagerDuty triage and alert-enrichment tools to reduce investigation time.
Mentored Dublin and Prague engineers, supported on-call readiness, and contributed to hiring for the Dublin team.

Cloud Native Engineer · Container Solutions August 2019 - August 2020

Client-facing consulting on Kubernetes, Istio, CI/CD, automated testing, deployment strategy, and cloud-native operating models.
Worked with engineering teams to make delivery and operations more reliable and repeatable.

Senior Developer · Marketplacer November 2016 - June 2019

Led development on major client implementations, platform scalability, and data integrations.
Moved deployment work into CI, improved deployment visibility, and significantly reduced build times.

Senior Developer · Vinomofo August 2016 - November 2016

Built warehouse and delivery integrations; fixed critical fulfillment logic with test coverage; resolved database locking and transaction issues.

Senior Dev/Ops & Lead Roles · Netfira Pty. Ltd. 2012 - 2015

Progressed through senior development and Dev/Ops leadership roles; managed AWS infrastructure, mentored developers, led hiring, and worked closely with product stakeholders.

Early Career 2006 - 2012

Moodle HQ: core Moodle development, planning improvements, mentoring, integration testing. Future Health Media & Technology: co-owner/ops/dev for HealthEngine and Popfossa. Buzka: junior developer to development lead/CTO.

Technical Strengths

AI engineering: LLM tooling, RAG, MCP servers, agentic systems, red-teaming, evaluators, tool orchestration.

Real-time systems: voice UX, streaming pipelines, latency budgets, failure handling, observability.

Infrastructure: Kubernetes, Docker, Pulumi, GitOps, AWS, GCP, bare metal, Grafana, Prometheus, Loki.

Languages: Ruby, TypeScript, Python, Go, Crystal.

Gerry Caulfield

Hey, I'm Gerry

Currently building Sandegg, a multilingual voice calling product

15+ years building software. Picked up the infra side after being told once that something couldn't be done. Turned out I love both, and they sharpen each other.

Low-latency Tool orchestration Observability Evaluators & Guardrails

Sandegg: real-time multilingual voice calls in private preview. You speak yours, they hear theirs.
Local-first personal assistant that outperforms Siri/Google/Alexa on context continuity, cross-session memory, vague queries, and self-adding tools.
SRE/backend history: AWS/GCP/Bare Metal; Kubernetes/GitOps/Pulumi; Grafana/Prometheus/Loki; cut deploys from hours to minutes and raised SLO adherence.

What I build

Tool-calling agents that actually handle failures gracefully (with timeouts, retries, and proper error handling)
Smart planners that break down complex tasks while respecting latency budgets
Retrieval systems optimized for short context and predictable costs
Guardrails & evaluators that catch issues before your customers do
Streaming UX that feels instant.

What I'm building

In production + active R&D

Sandegg, multilingual voice calls

Real-time multilingual voice calls. You speak yours, they hear theirs. In private preview at sandegg.com.

Sandegg multilingual voice call interface showing a live translated Cantonese caption

Demo

The first end-to-end call was with my brother-in-law in Shanghai. He speaks Cantonese, I was in Dublin speaking English. Neither of us speaks the other's language. We just talked.

More recently, my sister in Galway (English only) had a full conversation with my wife in Cantonese. They laughed together. The laughter landed in time. That's the real bar.

That's sandegg.com.

Status

Private preview. Letting people in a few at a time.

Local-first Personal Assistant

A powerful desktop voice assistant with persistent memory, self-tool generation, and deep OS integration. Built to outperform mainstream assistants on contextuality and actionable help.

Problem

Speech is about ~3-5 times faster than typing. Being able to trigger voice typing with a wake word and switch to an intelligent assistant is something I've wanted ever since seeing the movie Her. I'm not quite there yet but it is shockingly close.

Capabilities

On demand: The system uses Voice detection, commands and wake words to know when to listen and to what
Cross-session recall: Long-running conversational memory;
Self-extends: The agent can write and register new tools for its own use
Search: Web + local docs in-context
Desktop control: reads clipboard; types into fields;
Desktop context WIP: reads from apps across my desktop (hard part already done)

Demo

This was my first ever YouTube video, I'll improve and gain confidence talking to myself on camera, but for now please bear with it. :) I noticed people stopped watching before even seeing the interesting part, so I've now set this to skip right to it. Feel free to go back if have a higher tolerance for rambling.

Why it Matters

Beats Siri/Google/Alexa on context handling, adaptability, and actual task execution — while running locally for privacy and low latency.

Where I've been

Founded sandegg.com, multilingual voice calling

Solo • August 2024 - Now

Built sandegg.com, a real-time multilingual voice-calling product in private preview. Own the product, voice pipeline, latency work, deployment path, and early user testing.

Side R&D during the same period:

Personal agentic assistant with persistent memory and self-extending tools (shown above)
Mobile apps + on-device design system for live LLM iteration
Geospatial visualization tooling

Senior Site Reliability Engineer

Workday • August 2021 - 2024

Supported global applications in data centers, GCP, and AWS Kubernetes clusters. Focused on automating operations to simplify management of customer-impacting issues and enhance recovery times.

Key achievements:

Improved internal CI/CD systems for company-wide use
Developed tools for automatic triage of PagerDuty issues and enhanced alerts with relevant context to reduce investigation time
Improved monitoring and alerting systems across global infrastructure
Managed interviewing and hiring for the Dublin team
Mentored team members and prepared Dublin and Prague teams for on-call responsibilities
Collaborated daily with US-based teams, contributing to the successful establishment and growth of the Dublin team

Cloud Native Engineer

Container Solutions • August 2019 - August 2020

Client-facing consulting work improving how organizations build and deliver software using cloud-native technologies.

Key achievements:

Rolled out CI/CD systems using Kubernetes, Istio and cloud-native tooling
Mentored engineers on automated software testing, deployment strategies and team effectiveness
Worked with highly skilled engineering teams to make operations run smoother
Focused on practical implementation of cloud-native patterns

Senior Developer

Marketplacer • November 2016 - June 2019

Development, platform scalability and data integrations for two of the biggest clients on the platform.

Key achievements:

Led development on major client implementations
Mentored other developers on complex ecommerce challenges
Transitioned deployments into CI pipeline to remove mental burden from developers
Improved visibility into deployment process for entire team
Significantly sped up CI build times through optimization

Senior Developer

Vinomofo • August 2016 - November 2016

Key achievements:

Development of new warehouse and delivery integration
Fixed critical legacy fulfillment logic and added comprehensive testing
Discovered and resolved tricky database locking, transaction and timing issues
Added audit trails for critical business processes
Updated site design for national ad campaign with automated launch
Extracted common code into reusable components

Senior Dev/Ops & Lead Roles

Netfira Pty. Ltd. • 2012 - 2015

Multiple roles with increasing responsibility: Senior Developer → Lead Web Dev/Ops → Senior Dev/Ops (External)

Managed and improved AWS infrastructure
Trained and mentored other developers
Led interviewing process for new team members
Developed new infrastructure and projects
Worked extensively with product stakeholders

Early Career

2006 - 2012

Moodle HQ (2011-2012): Core Moodle development, planning improvements, mentoring, integration testing
Future Health Media & Technology (2009-2011): Co-Owner/Ops/Dev - Product planning, infrastructure, development for HealthEngine and Popfossa
Buzka Pty Ltd (2006-2009): Junior Developer → Development Lead → CTO - Full career progression at technology startup

The story so far

I started out as a developer and have been one ever since. Picked up the infra side after being told once that something couldn't be done. Turned out I love it just as much, and the two disciplines sharpen each other.

What I'm building now: A multilingual voice-calling product (in private preview at sandegg.com) and a local-first personal assistant. Both sound simple until you try to make them reliable enough for daily use.

Toolkit

AI Engineering: LLM tooling, RAG, red-teaming, MCP servers, agentic systems
Real-time Systems: sub-second latency at production scale
Infrastructure: Pulumi, Kubernetes, Docker, multi-AZ AWS/GCP
CI/CD: Sub-30-second builds, parallel testing, deployment automation
Languages: Ruby, TypeScript, Python, Go, Crystal
Leadership: Mentoring, hiring, team building, on-call preparation

Let's chat

I'm always up for interesting conversations about AI engineering, reliability challenges, or just geeking out about tech. If you're building something cool and think we might work well together, I'd love to hear from you.