Engineer · Speaker · Guitarist
Right now I'm a Forward Deployed Engineer at Quarterzip AI, where we build customer onboarding and activation software for B2B teams. Before that I was at Redactive AI working on semantic security for enterprise AI, until Recordpoint acquired the team.
Earlier in my career I was a Cloud Architect at Amazon Web Services, designing public-sector solutions for universities, government, and not-for-profits — and before that a Software Engineer at Localz, building last-mile delivery software for retailers around the world.
I'm a Monash graduate in Electrical Engineering, Computer Science, and Maths. While I was there I ran wired, the IT society, as president in 2017, and helped organise UNIHACK Melbourne — student hackathons across Australia.
Through 2025 I served on the national committee at Young Engineers Australia, after co-chairing the Victorian chapter — both roles spent running events for engineers early in their careers.
Outside work, it's running and bouldering around Melbourne, plus gaming, karate, biking, and tinkering with electronics — at least one of those most days. If you want to see what I've built, my projects are below and my resume is a click away.
Best by email — calendar second, ravens third.
Drop a line at robert@kochie.io or .
Opt-in only — unsubscribe at any time. Powered by Beehiiv.
Badges I've collected, via Holopin.
Some shipped, some still in progress.
Adds missing metadata to photos exported from Facebook so they import into photo apps with the correct date and location.
Open-source typing tutor for Linux, Windows, and macOS — improves your typing skills without nagging you.
Technology demonstration of a contact-tracing application built on the AWS Cloud Development Kit.
A small webapp that follows the mouse.
A real-time n-body particle simulator for the web.
A trading platform for crypto-currencies.
A benchmarking tool for JavaScript projects.
Final-year thesis for Electrical and Computer Systems Engineering.
Final-year thesis · October 2017
Convolutional networks are too heavy for small devices. This thesis modifies their activation layers so models are smaller and faster to run on embedded hardware, and measures what it costs in accuracy.
DDD By Night
Our company built an AI product and released it to customers — an immense undertaking that taught us a lot about taking an idea from prototype to the real world. The technical and business challenges, and how we solved them.
Programmable
With generative AI tools rapidly integrating into every corner of the enterprise, security risks are evolving faster than ever. As developers, you're on the frontlines of this revolution — tasked with safeguarding sprawling, unstructured data that traditional security methods can't handle.
We'll reveal the hidden vulnerabilities that GenAI exposes — accidental data leaks, misconfigured permissions lurking in knowledge bases — share cutting-edge security strategies, and introduce tools that can fortify your data defenses.
How to Build Ones That Actually Drive Business Value
NDC Sydney
AI agents are everywhere - but most of them don't work. They drain engineering time, frustrate end users, and burn budgets without ever proving their worth. The result? Shiny demos that impress in the boardroom but fail in production.
This session shows you how to flip that script. I'll share a practical framework for designing and deploying AI agents that create measurable business impact, agents that align with real workflows, integrate securely, and deliver results people actually care about.
If you want to stop wasting resources and start building AI agents that stick, this talk will show you how.
Lessons from building FDE at a startup
Upcoming
The hardest part of building a startup is finding product-market fit while inventing the business around it. Forward deployed engineers fill the gap between what your product does and what customers actually need, and in AI, that gap is where deals die.
We built an FDE team and got most of it wrong first: a single word in our positioning was teaching customers to churn, deals closed without anyone owning deployment, and my team sat squarely in the crossfire between engineering and sales.
This talk covers what broke, what we fixed, and how FDEs can use their position in the middle to shape both sides - so customers get solutions, not handoffs.
In this session learn how to build Retrieval-Augmented Generation (RAG) pipelines for AI agents. We'll explore the principles of what looks good, and how to break down complex business problems so they can be reliably solved by RAG assisted AI Agents.
In this session we'll start from the basics and discover what RAG is, we'll then create a simple pipeline into a Large Language Model, finally we'll create a multi-modal RAG pipeline that can handle complex enterprise security requirements.
Share on X ↗Python is slow… until it isn't. This talk shows a no nonsense playbook for squeezing 10x more performance out of real Python code without rewriting everything in C. Through live demos we'll profile, vectorize, batch, and compile hot paths turning sluggish scripts and services into production-ready speed demons. You'll leave with a clear checklist and the confidence to know which optimization matters most in your app.
Share on X ↗What if you could tap into satellites orbiting the Earth right now - without a billion-dollar budget or your own dish farm? Thanks to cloud-connected ground stations, anyone can capture and stream downlink data directly from satellites, process it in real time, and feed it into modern AI pipelines.
In this talk, we'll go hands-on with the growing ecosystem of cloud ground stations and show how to transform raw space signals into usable datasets for computer vision, climate science, and machine learning.
Whether you're an AI engineer, data scientist, or just a space geek with cloud credits to burn, this session will show you how to bring the view from orbit straight into your ML models.
Share on X ↗Badges I've collected, via Holopin.
Posted to blog.kochie.io.
I’m not sure why this still needs repeating, but please—stop writing recursive Lambdas.
An in depth guide for using Apple's StoreKit with Electron
Because anything is better than the train-wrecks we have at the moment.
