# Ryan Hay

> AI engineer in Sydney building production LLM agents. Core engineer on Invent, Relevance AI's agent that builds, edits and debugs customer agents through MCP tools — shipped from prototype to general availability, now used by 750+ users across 640+ organisations a month. Works across multi-agent runtimes, agent permission models, LLM evaluation and AI-native engineering workflows (Claude Code). Studying Actuarial Studies / Computer Science at UNSW; multiple hackathon wins.

- Location: Sydney, Australia
- Current role: AI Engineer at Relevance AI
- Contact: hey@ryanh.co
- Availability: Happy at Relevance AI, but open to hearing about exceptional AI engineering and founding-engineer roles. Based in Sydney; open to roles in Sydney or remote.
- Now: AI Engineer @ Relevance AI. Studying ACTL/CS at UNSW, winning some hackathons, always exploring ideas on the side. Based in Sydney, Australia.

## Experience

### AI Engineer — Relevance AI (Series B AI agent platform) (Mar 2025 – Present)

- Core engineer on Invent, Relevance's AI agent that builds, edits and debugs customer agents through MCP tools, in a two-person team. Shipped it from prototype through a ground-up rebuild to general availability (Aug 2026); it now serves 750+ users across 640+ organisations a month, including 50+ enterprise accounts, who have built 1,900+ agents and 2,900+ tools with it.
- Built its multi-agent runtime: parallel subagents with guardrails on destructive actions, memory and recall across conversations designed around prompt caching, scheduled check-ins with escalation, and access to customers' own MCP servers.
- Designed and built the permission and approval model that lets enterprises control what the agent can do on a user's behalf (per-tool risk classes, user- and conversation-scoped grants, org-level model policy), now governing 37,000+ agent actions.
- Built and owns Invent's LLM evaluation suite; rebuilt it around behavioural rubrics validated against 218 hand-labelled runs, cutting false passes from 53% to 11%.
- Drives AI-native engineering across the team: internal Claude Code plugin marketplace, automated review agents encoding team review standards, isolated per-worktree dev environments with automated browser QA, and an agent-friendly CLI for the platform.
- Built the platform's analytics product (1,100+ users across 1,000+ organisations each quarter) and an SQS-based notification system for agent alerts and escalations over Slack and email.
- Shapes Invent's roadmap each quarter from customer research; member of the production on-call rotation.

### Software Engineer (part-time) — Seventh Beam (Mar 2022 – Mar 2025)

- Built the company's first proprietary product while in high school: a financial-transactions platform migrated from a legacy Kotlin back-end to Dart/Flutter on Google Cloud, handling 100+ transactions/s.
- Built C# API endpoints for Donut King's 'Occasions' ordering system, integrating Microsoft Azure, TASK Software and Uber across 8+ locations.

## Awards & Speaking

- Speaker, "Claude Code Isn't Just a Coding Tool", developer event hosted by Vercel (2026)
- 2nd, ElevenLabs x a16z AI Agents Hackathon (2025)
- 3rd, Startmate Fellowship Challenge S25 (2024)
- 1st & People's Choice, DevSoc Blueprint Hackathon, 110+ participants (2024)
- 1st, Relevance AI Tools Hackathon (2024)
- 1st with Groq, Eucalyptus x Build Club Hackathon (2024)
- ADF Future Innovators Award (2023)

## Education

- UNSW Sydney: Bachelor of Actuarial Studies / Computer Science, expected 2028 — WAM 77 (Distinction)
- Rose Bay Secondary College: ATAR 99.15 (top 0.85% in Australia), 2023

## Skills

- AI: LLM agents, Multi-agent systems, Model Context Protocol (MCP), LLM evaluation, Prompt caching, Claude Code, Anthropic API, AWS Bedrock
- Languages: TypeScript, Python, SQL, C#, Dart
- Technologies: Node.js, React, Vue/Nuxt, PostgreSQL, MongoDB, Redis, AWS (SQS, SNS, Lambda, DynamoDB, CloudWatch), Google Cloud, PostHog

## Projects

- quackerscroll (hackathon, winner, 2025-dec): TikTok + Tinder for recruiting technical talent. Lyrathon S1.
- rubberducky.me (hackathon, 2025-nov): Speak your system architecture and diagram it with a voice agent that simulates loads. ElevenLabs Worldwide.
- gifty.ai (hackathon, 2025-mar): Voice-activated AI agent to guide you through choosing the perfect gift. UNIHACK.
- agility.ai (hackathon, winner, 2025-feb): AI agent that conducts 1-on-1 standups, integrating with Jira and GitHub. ElevenLabs x a16z.
- skillscale (hackathon, winner, 2024-dec): Streamlined tech recruitment process. Startmate Fellowship Challenge.
- doordash-agent (hackathon, winner, 2024-dec): Demo-able integration of DoorDash's API with Relevance AI. Relevance AI Tools Hackathon.
- receipt-analyzer (hackathon, winner, 2024-dec): Take a photo of your receipt, get a nutrition summary and healthier shopping list. Eucalyptus HealthTech.
- lectrify (hackathon, winner, 2024-oct): Real-time quiz platform handling 65 concurrent users. DevSoc Blueprint.

## Journey

- Nov 2025: Top 1% in Valorant. Also host gaming tournaments on the side.
- Mar 2025: Joined Relevance AI. AI Engineer — Left Seventh Beam after 3 years to join Relevance AI as an AI Engineer. Powering a builder platform for thousands of enterprise users.
- Dec 2024: Won prizes at 3 hackathons. Feels like we finally understand the game, Came 3rd in Startmate's student fellowship program, won over $2k in prizes with Chris Yoo.
- Oct 2024: First hackathon win. Won DevSoc's Blueprint with Lectrify, a real-time quiz platform handling 65 concurrent users built in 5 days.
- Feb 2024: Started at UNSW. Computer Science + Actuarial Studies.
- Dec 2023: Graduated with 99.15 ATAR. Top 0.85% in Australia. Won the ADF Future Innovators Award.
- Mar 2022: First engineering job. SWE @ Seventh Beam — I was 16yo interviewing for my first real internship.
- Jan 2022: Top 1% on TryHackMe. A 3 month sprint into cybersecurity reignited my passion for building after a year exploring other interests.
- Apr 2021: Moved to Australia. Started Year 10 at Rose Bay. Took a step back from programming to pursue other interests.
- Mar 2020: Helped grow Sector's Edge 1k -> 10k members. COVID hit, found an indie game community. Helped with outreach, ran the content creator program. Learned that community building matters as much as the product.
- Nov 2019: Front page of r/python. An optimization algorithm that recreated images using randomly placed triangles. Stayed on the front page for a week. It validated I could build something other people thought was cool.
- Jun 2019: Built neural network from scratch. built with python using only standard libraries. Created a working MNIST classifier at 13.
- Jan 2019: Innovation & Technology Scholar. Youngest recipient at my school of 2,500+ students. This gave me confidence I could innovate.
- Jun 2018: Moved to Dubai. Built Ultimate Squash in Scratch, a full game with upgrades and perks. It was probably my first project that 'Launched'
- 2015: Started programming. Minecraft command blocks and Khan Academy JavaScript. Showed my dad what I built, he encouraged me to keep going. The feeling of 'speaking things to life' hooked me.

## Links

- [Website](https://www.ryanhay.com): personal site
- [Resume (PDF)](https://ryanh.co/resume.pdf): full resume
- [LinkedIn](https://www.linkedin.com/in/dmryanh/)
- [GitHub](https://github.com/TyperOfCode)
- [X](https://x.com/rynhay)
- [Email](mailto:hey@ryanh.co)