QCon London

Pikiv developers at QCon London 2024

Conference

QCon is one of the world’s leading conferences for senior developers, architects and DevOps-engineers. Piktiv sent four of our developers to Westminster, the heart of London, to discover the latest trends and future prospects of the IT-industry.

QCon London 2024 hosted six parallel tracks of talks for three entire days. Being four people, we decided to split up and try to gain as much knowledge as possible (in-between the excellent food and snack servings, and views of the British parliament).

Erik’s highlights:

EventStore DB’s talk about Event Sourcing at the database level felt like a perfect fit for many projects that I have worked on. I especially liked their focus on business events, rather than data-events (for instance: “customer moved”, “customer update phone number” vs “customer data updated”)
The presentation about zero downtime PostgreSQL migrations using pgroll, and how they used it to migrate data across different environments.
Webassembly on the server, presented by SPIN! Orders of magnitude less memory and CPU usage in comparison with traditional Docker images. And also architecture agnostic (I never want to have to cross compile docker images again…)

Jesper’s highlights:

The talk about how ClearBank went from a start-up to a scale-up while maintaining the speed and autonomy of a smaller company was very inspiring and showed that even in a heavily regulated industry bureaucracy can be kept to a minimum.
Eb Ikonne gave a really nice talk about how to go about making changes in the organisation you are working in regardless if you have a formal leadership position or not. Especially highlighting the power of small incremental changes.
Being a Bad Influence - this talk went through some of the potential pitfalls in people management and had both good information as well as some very relatable stories.

Joe’s highlights:

What can you learn from the fastest code in the world, presented by Alan Elder. It is an interesting presentation on the cloud. What can you possibly do to reduce the latency and make the processing speed super fast. From changing the data structure to pre-loading data, allocate data to different layers of Ram, increase batch sizes to squeeze efficiency of each cycle. But, everything comes with costs. In this case the readability of the code and error handling downsided.
How to send 8 million messages in 5 seconds, from Duolingo. I think it is also an interesting presentation in which they want to send notifications to their users with their tv advertisement for the Super bowl. The technical side of Aws serverless event driving approach was easy to understand and the scalability was achieved by preplanning, preloading, and in advance warming up. In addition the communications between Aws, and testing part was quite important to ensure a critical event can be implemented with timely precision.
Postman with new AI features.
Postman presented their features on AI generated testing. It can also write documentation of APIs. Postman can suggest some tests and also provide a way the user can create test titles and it would generate tests based on the test title. These could be an efficient way to maintain the documentation and create api tests. But, it could has some AI related impacts such as data security and possible errors.

Fredrik’s highlights:

In the first talk of the AI track Jay Alammar from Cohere gave a great intro on the state of Large Language Models and what techniques are the latest to be widely adopted. Getting to learn more about Retrieval-Augmented Generation and the structures LLMs can function and flourish in was very neat, especially as several other talks followed up on these techniques in more detail.
Michael Brunton-Spall from the UK Cabinet Office held an interesting talk on modern security breaches. He walked us through a few different real life examples of exploits using unexpected attack targets on low prio systems to gain access to high prio systems. This was complemented in a later amusing talk where Rich Smith talked us through how he breached a "safe code storage" service by feeding its LLM AI assistant a limerick.
Gregor Hophe from AWS shared his thoughts on being a system architect in his talk Thinking Like an Architect. His perspective on a system architect's role in communication between departments/people with very different needs, wants, and competences provided some neat insights.