Insights that unlock opportunity: Top takeaways from LeadDev London

At the end of June, I attended LeadDev’s conference in London with more than a thousand senior development professionals. Karat was a conference sponsor, and I was joined by a cohort of my UK-based colleagues who ran our booth in the sponsor hall.

LeadDev London conference attendees visit Karat’s booth to learn more about the Interviewing Cloud.

I was there to speak about hiring processes with a focus on how to make them more equitable and efficient, but most importantly, how to do the first in a way that will help achieve the second. But, I was also there to attend the conference, to listen to different perspectives on what’s happening in the industry, and participate in the conversation — and there was a lot going on in the conversation!

A note of gratitude for LeadDev organizers

First, I’d like to thank the folks at LeadDev for helping us arrange conference passes for a group of folks who make Karat interviews possible. Four interview engineers were able to join us at the conference with general passes. They attended talks, met with other sponsors, and enjoyed the after-parties.

An incredible speaker experience

As the conference got under way, I was given the opportunity to take the stage to share insights on how to “engineer a more equitable hiring process.” Delivering an equitable experience is just as vital to candidates as it is to companies. I shared some best practices we’ve seen work over thousands of interviews, as well as some key next steps that engineering managers can implement now. Watch my full talk from LeadDev London now:

I’ll share my take on many of the other excellent talks from LeadDev London in a follow-up post. More details to come soon!

A focus on AI

This year, there was a lot of discussion about AI. I had the most engaging conversation about it with the folks at SourceGraph, who have been working to more directly enable a large language model (LLM) to operate with specific context about a given codebase. Their project, Cody, is interesting because it acknowledges some of the limitations of general model training. Its model also indicates that some of the actual effort to make LLMs useful as tools has begun.

Given Karat’s role in the interview process, we’ve been looking a lot at how people might use LLMs in the day-to-day work of software development. Don Gannon-Jones and Natalie Wright shared some thoughts in an article Karat drafted for LeadDev. You can learn more about how AI tools can help your company with its interview processes here.

A deeper dive into AI & test case development

In a different vein, I also chatted about AI with someone from QA Wolf, a third-party test case development and maintenance company. I have a lot of connections in that field from my experience working as a test engineer, and I’m seeing a lot of excitement about the prospect of using LLM systems to generate test cases. But, when I asked the folks from QA Wolf about where they’re using these tools, they pointed out that generating test cases isn’t actually the expensive part of owning and maintaining a test suite. The bulk of the effort goes into reacting to test failures, and triaging whether the failure corresponds to a real issue in the product. If we’re going to get automated support in testing, we need to focus the tools not on what’s easy to automate (i.e., the test-case generation), but on what’s expensive to do manually (i.e., actually dealing with test case failures).

The overall theme from LeadDev London

Many of my conversations at the conference reinforced that point: looking at what new technologies can do is interesting, but often the real value is unlocked not by finding a way to apply a new technology, but having a deeper understanding of a problem and applying well-understood solutions to it more effectively. As much as generative AI is fascinating, and has led to some flashy demonstrations, we keep seeing the approach of humans and tech working together making real improvements.

I also had a great chat with the folks at Skiller Whale, who are applying that same human plus tech approach to technical learning. They’re helping connect developers who need to learn something with mentors in one-on-one and small-group environments, based on the observation that the best learning happens with a mentor, but not every senior developer is also a good teacher. There’s an obvious parallel between what they’re doing and how we conduct interviews: A human with the right skills and training, supported by good tools, is more effective than the best automated systems we can build.

I came away from the conference reminded of just how essential humans are to the entire endeavor of building technical systems. While we may be at an inflection point in the capabilities of language-modeling systems, and we may see new models of interaction open up in the next few years, we’re not removing the need for a human hand in building our own future. If we build it right, we might simply be opening up the field to allow more humans in, by reducing some of the barriers between understanding a problem and building tools to help solve it.

An invitation to join us at LeadDev West Coast in October

The opportunity to participate in and join LeadDev London was invaluable, and I think LeadDev West Coast will be just as invaluable this fall. Be sure to join Karat in the U.S. in Oakland, California from October 16–17, 2023. Get a discount on your conference pass with the promo code KARAT when you sign up here.

Stay tuned for a summary of all of the best talks from LeadDev London right here on Karat’s blog. In the meantime, I hope to see you in Oakland in October!