Presentation: Front End Architecture in a World of AI

Track: Developing/Optimizing Clients for Developers

Location: Majestic Complex, 6th fl.

Duration: 11:50am - 12:40pm

Day of week:

What You’ll Learn

  1. Find out about AI and how to improve the solution engine based on human responses.
  2. Hear how AI engines can interact with machines.
  3. Learn that state in large scale web applications can be distributed.


Increasingly humans are no longer the only clients interfacing with your data, machine and AI clients are starting to manipulate data in real time and can create suggestions or influence human behaviour. This talk deals with the complexities of dealing with many different clients that are not all human and how to build a single interface that both humans and AI clients can leverage.

Using state container methodologies like redux we can treat a chain of machine/AI data manipulations the same way as a real person interfacing with your app. You can look at it as a headless version of your front end that cuts development time and allows for a testable, predictable, flexible and scalable architecture.

We're going to be dealing with topics like state synchronisation methodology, time traveling, comparing OT (Operational Transforms) vs CRDT (Conflict-free Replicated Data Types), vector clocks and merkle chains in javascript.


What is the focus of your work today?


Right now I'm trying to create a scalable front-end for a very fast growing platform that's used to solve a wide variety of modern manufacturing problems. We're trying to support both humans and machines, and allow enough flexibility to scale while still adding new features constantly. So we're trying to create a simple UI that's also progressively revealing complexity because manufacturing problems are notoriously difficult as they progress. At the same time this is all done in the browser because we want to do fast iterations. We want to keep our development cycle very fast, and so we have to support 3D in the browser, we have to support multiple complex workflows. There's a whole set of challenges that are my current focus.


What's the motivation for the talk?


One of the big challenges in the AI is the feedback loop. AI when it shines you can advise users to solve this specific problem, but it only really works when the user can actually react according to the advice that AI gives. Given the ability for the AI to learn from the user while at the same time making progress on calculating suggestions can be a very hard problem to deal with. And my experiences over the past two years might be insightful for others facing similar issues.


What would you describe as the persona and level of the target audience of your talk?


Mainly senior software architects type level, people with a good understanding of current front-end architecture and frameworks that are being used in the field right now. And just general people that are interested in more advanced complex workflows, not your average WordPress websites.


What do you want the audience to walk away with?


Distribute state is possible, but just be aware of the different approaches that have their pros and cons.


In the abstract, you talk about non-human uses of software. What examples can you give me of non-human software users?


By non-human users I mean for example AI clients which want to preemptively or proactively calculate certain values to then better accommodate or better make better suggestions to the user. In our use case, we have a workflow where we're trying to create a turbine blade which consists of about 1000 parameters, the user going through every single parameter would cost a couple of days worth of engineering efforts whereas an AI agent based on previous workflows and previous experiences might suggest or might create a UI and structure UI that's different for the user and help the user go through the process faster.

Speaker: Thijs Bernolet

Front End Architect @oqtonai

At the age of 7, Thijs wanted to become an inventor. It didn’t matter what, he just wanted to invent things. Several broken washing machines, a fried television and 8 blown amplifiers later he saw a magical light shining from a machine they called a ‘Macintosh’. He’s been hooked ever since. Thijs is always looking for new ways to interact with machines and everything involving technology. After learning every programming language within his reach, he choose an experimental education at the Royal Academy where he got a strong focus on concept and creativity. He later co-founded a company called We Work We Play where he further experimented with 3D in the browser, mobile and interactive installations. He also cofounded Nerdlab in Belgium, a non-profit organization that wants to show kids how to hack cool things together with hardware and software. They recently send an iphone to space and build a Ghostbusters replica backpack with GPS that gets carried around Europe. After joining and getting aquired by Autodesk, he moved to the Bay Area to continue work on the real time browser breadboard simulartor and later on led the development effort of the new Autodesk design system (HIG). He then moved on to join Oqton, a new startup that aims to revolutionize manufacturing by creating an operating system for the factory that is driven by AI.

Find Thijs Bernolet at

Similar Talks