connor.fraser-grant.com
Connor
Fraser-Grant
{ Software Engineer — AI Tooling & Infrastructure }
About
No university degree. No formal qualifications. Started freelancing, taught myself, worked my way into startups. That path isn't a gap — it's the whole point. I learn by building, and I've built a lot.
Worked at Functionly alongside a world-class team, building a tool for organisational design. That's where I built my first AI agent feature and got deep into how AI actually works in a product — not in a demo, in production.
Now I run Fraser-Grant Agency, building AI tools and infrastructure. Two main projects: OrgSDK (multi-agent code execution platform) and Quokly (AI research platform focused on citation accuracy). I'm looking for opportunities to unlock AI usage in organisations — custom tooling that leverages what's actually capable right now.
"Augmentation-first. AI is a tool that will unlock a lot, but shouldn't be assumed to replace people. Optimistic that we might get to something more transformative, but it's a long way off. The opportunity right now is leverage — new tools to run faster, improve quality, capture opportunities."
Work
Functionly
Organisational design tool — helping leaders map roles, positions, and relationships that form a business.
Read more
Problem
Org design is complex. Leaders need to visualise and restructure how their organisation actually works — roles, reporting lines, headcount, the lot.
Approach
Joined as a software engineer after freelancing. Built the AI agent feature from scratch — the first real exposure to shipping AI in a product. Got deep into how organisations work: how roles and positions and relationships form the backbone of a business.
Result
Learned from a world-class team. Saw the evolution of AI happen in real-time. Built an AI feature that real users interact with. It's where the AI engineering path started.
OrgSDK
Multi-agent code execution platform — AI agents that run TypeScript workflows in isolated VMs.
Read more
Problem
AI agents need a sandbox. Somewhere to execute code, access tools, persist state, and coordinate with each other — safely and reliably.
Approach
Building the harness from scratch. Agents get isolated VMs, persistent Postgres state, a plugin system (Google Workspace, databases, LLMs), scheduling, approvals, and orchestration. TypeScript throughout.
Result
In active development. The foundation is solid — agents can execute workflows, manage state, and coordinate. Now building toward production use.
Quokly
AI research platform for the Australian market — accurate, well-cited research reports.
Read more
Problem
Most AI research tools hallucinate. For the Australian market in particular, accuracy and citation matter — you need sources you can verify.
Approach
RAG-based research with citation accuracy as the core differentiator. Not generating plausible text — generating verified, cited, useful research output.
Result
In beta. Focused on getting the citation accuracy right before scaling.
Path
Founder & Engineer
Fraser-Grant Agency
Solo practice building AI tools and infrastructure. Created OrgSDK (multi-agent execution platform) and Quokly (AI research platform). Taking on AI-focused client work — custom tooling that ships.
Software Engineer
Functionly
Built the AI agent feature from scratch for an organisational design tool. Worked alongside a world-class team. Got deep into AI engineering, org design, and shipping features that real users depend on.
Freelance Engineer
Self-employed
Taught myself software engineering by building real things for real clients. Full-stack — deployment, CI/CD, backend, frontend. No degree, no shortcuts. Just shipped until someone paid me to ship more.
Approach
Practical AI. No hype.
Full-stack from day one — I understand the whole system, not just one layer
AI as augmentation, not replacement — leverage what works now
Ship early, iterate fast — daily visible progress on everything I build
Self-taught pragmatism — I learn by building, not by reading about building