Doing DevRel In the era of GenAI

Uttam: Really nice to be here. I have actually been at a number of DevRelCon events in the past, both as an attendee and and, as a as a speaker, and I have learned so much from the community. Again, like, this is one of those, trades which I call as a practice. We are always trying to learn. And as as a lot of other, professions where you are practitioners, it's also important to be open and sharing with others and learning others tactics and and and and ourself moving in that direction.

Qualcomm's Shift to Developer Relations

I'm gonna break my talk into two broad parts. One of them really talks about why is Qualcomm here. You might have not seen a lot of people from Qualcomm at places like DevRelCon or working with third party developers Because historically, being a technology company, we had a a certain, market segment that we were working with, but but that's expanding. So part part of the talk is gonna be why Qualcomm cares about third party develop. Important for us to share that with you because as we are building our DevRel team, we are hiring.

The engineering teams are hiring. So if if that piques anyone's interest, come and talk to me. We also have Paul Torres here, who's who's my colleague, from San Diego. Come and talk to one of us, and then hopefully, that can also lead to some of the opportunities within Qualcomm that that can work out. Right?

And then the second talk is very specifically and tactically, as we are building this new function, within Qualcomm, of DevRel, how are we using GenAI in that journey? Like, practice what you preach, taking that principle of dogfooding to heart. And it's been very useful for us in the current environment with tight budget situations, staffing being how can you really get more out of what is offered to you from a from a DevRel perspective? So this is something that you all probably have seen once or or or more in the last eighteen to twenty four months. Still, remember, I think it was twenty twenty twenty three, right at the beginning of the year, chat GPT screenshots started getting viral on Twitter.

And then that was the inflection point when suddenly the technology that actually my previous employer, Google, had had worked on transformers was starting to get productized. And that kinda started this, what I'm calling as the fourth industrial revolution, which is driving this AI change going forward. And we all had heard, read about AI algorithms and and the the concepts and the theories in the last couple of decades in textbooks. What is really different and what is driving these changes, now compute is available to really take some of those ideas and and make them practical. And also with the advent of big data, so you you take all of those algorithms and and concepts and theories, marry it with compute power that's available right now, and thanks to big data, suddenly all of these solutions start to emerge.

And and I think it's only getting started. It's it's really gonna transform the way we live our lives. It's really gonna transform the way humans have, have lived on this planet, and and I I think we're gonna see those massive changes going forward. But as we have seen in the last couple of technology waves, especially mobile that I was able to see, and close with Android, The large companies provide the platforms. The actual use cases come from developers.

How are how is this technology gonna be implemented in a suburb in New York City? How is this technology gonna be implemented by by cities and towns, in in, say, Africa and parts of Indonesia and India? So that's why this phrase of developers build a future is critical. And, of course, as we heard in the morning, what it takes to be a developer is also changing rapidly. You you can start programming and and building solutions using natural language.

You don't have to be an expert in in in programming languages, which really opens up the the field for a lot more. So GenAI, of course, is is being really critical there. But from a Qualcomm perspective, and this is where it starts to get interesting for us and we get involved, is on device AI. So, of course, like, cloud is powerful, and you can run a lot of these workloads and and do training and also run inferencing using cloud. But then as AI gets ubiquitous and these solutions get deployed around us from things like robotics to Internet of Things, privacy, having more control to the data itself, making sure that the latencies stay low.

All of those factors become important. Sustainability, how much power is being consumed for running these workloads. And that's why on device is something that starts to get really important, and that's an area where Qualcomm, over the last three to four decades of constant investment in hardware, becomes, interesting for us. And and that's one of the reasons we are moving forward in this in this journey. So those of you who have not heard of Qualcomm, again, we've we are celebrating our forty year anniversary this year, a company that really has pioneered a lot of innovation when it comes to communication.

In fact, Qualcomm stands for quality communication. That's that's what the name stands for. But in the last few years, we have been moving in the direction of taking that communication power and and bring it to to compute as well. How do you bring, communication and comp compute power together? That has resulted in us powering the whole Android smartphone ecosystem.

Like, this advent of of GenAI and and some of the challenges that are coming up, which are very visible, the cost of inferencing, what we talked about privacy and latency on device is important. And that's where Qualcomm is moving forward, and and we are making some of our offerings available, for third party developers. So as you are thinking of deploying AI solutions around us, how can you use Qualcomm's technology in that journey? This is a visual representation of where we think eventually things will stabilize. So, of course, a lot of training starts on the cloud, Ease of deployment helps you get started quickly.

But then as scaling becomes important, as wide scale deployment becomes important, that's where you start to move in this direction. And and and there are, again, trade offs that that are important here, which is you need faster inferencing. You you need responses much, much quickly. Performance and efficiency becomes important. Immediacy, like, things that that need to live within a certain geographical location.

Of course, personalization. If you want to have AI give very personalized, responses to a user, a user may not want all the data to flow back to cloud all the time. They may want it to stay on that on that device. I personally think, eventually, things will stabilize in some sort of a hybrid setup where vast majority of use cases are being served on device, and periodically, you keep going back to cloud, to to pull any any major updates that are that are needed. Also, way AI is is moving forward and and all the agentic stuff that we are seeing, it's really being seen as a new UI.

Suddenly, text is is the way through which you interact with computers. Voice is the way through which you interact with computers. Photos is the way through which you interact with computers. Right? Whole app ecosystem that we have known for several years, gonna get transformed in a massive way.

And and that's, again, an opportunity for for new startups, new players to come in and get involved. Right? Like, don't worry about, okay, if someone already built an app and they are super successful and I'm trying to serve the same users, I can't get into that market anymore. This AI actually is is some sort of a leveler where the the UIs are more foundational, and, of course, all the back end data is is standardized. And through things like agent tech workflow, you can create new use cases.

So all of this is important for Qualcomm. And and from our journey perspective, what we are working on is offering tools to third party developers where you can build these edge AI on device use cases across industries. So going from automotive to industrial IoT to even compute, the laptop, business that that Qualcomm entered, last year to smartphones. How can we enable third party developers to start building these AI applications, really aiming for on device to start with, but then eventually also getting to the cloud stage and and and and hybrid. And one of the solutions as part of it is a product called Qualcomm AI Hub, which is a result of a startup, that Qualcomm acquired last year called TETRA.

So Qualcomm brought in TETRA, worked with the team, and launched this new offering called, Qualcomm AI Hub, which is literally a very easy way for you to get started if you are trying to build on device, applications. If you go to Qualcomm AI Hub, there are hundreds of pre optimized models that that run great on Qualcomm's Snapdragon platform. So you you literally pick one of those models, tweak it, test it on on our hardware, and and deploy it and get started. You can also bring your own model, which is a different path that is supported, and, again, like, go through that integration journey and and and and start, building solutions over there. In in just over a year, when since we launched, AI Hub, I think we launched it in February, over 1,500 companies are already using Qualcomm's AI Hub, and this number is literally growing as we speak.

And and across ecosystem, as I said, right, from, from auto to compute, to mobile, to to robotics. And then we're also working with large, hyperscalers, model makers to really give a wide range of choices for third party developers. So while this is a new area for us within Qualcomm, we we are optimistic about the progress while also being aware that there's a long way for us to go. We've also been building the ecosystem. So in in addition to AI Hub, Qualcomm also acquired, Edge Impulse, and and bring in a lot of their capabilities to really complete this picture when it comes to the ecosystem and and where things can go, where you can bring any of your data, use the foundational models either from Qualcomm's own AI research that's coming in or through Edge Impulse, then you can tweak the models, update them, go through either your own set of hardware or pick up some of the dev kits that Qualcomm is launching under our Dragon Wing brand, and then deploy it, and then that whole iterative process goes on.

So the point here being is this is a new area for Qualcomm. You probably would have known Qualcomm over the years as as a chipmaker for smartphones, but this is a new area which is critical for our, journey going forward. And that's where we are creating function from a from a developer product management perspective, but also DevRel. So a lot of DevRel people that you will see out there from Qualcomm are really trying to to serve third party developers who are building AGI applications. And as I said, like, we we're also hiring.

Leveraging GenAI for Developer Engagement

We're not hiring, like, a ton, but we're we're also hiring. So if it really piques anyone's interest, if this is an area that that excites you, come come and talk to us. Right? So while we have been observing this and and when I moved from Google to Qualcomm last year and literally trying to build a new DevRel function, it's also very obvious that the way we have done DevRel over the years will also get transformed in a massive way. Like, if all of these tools exist, that makes content creation, the way you analyze data, easy, and and we are talking about other industries and other professions being being transformed, why not DevRel?

It it's just a matter of time. And one way of looking at it is just get scared out of it. Okay. Like, I've done this job like this all my life, and suddenly, this AI is gonna come and and take over my job. The other way of thinking about this is I've always had less time than what was what I could have done.

So how can I use some of these tools to bring more efficiency into my day to day work and reinvest that time in relationships? Because at the end of the day, having that authentic conversation, being able to create and build relationship with your partners, with your with your developer community is critical, and I think that's the opportunity this this presents. So this is a typical funnel. People call it and name it in different ways. But from a DevRel journey perspective, you want your developers somehow go through this this path where, first, you want them to know that you exist.

That's the connect phase. Right? Then once you once they know you exist, you want them to start engaging with your content, with your resources, with your community. That's that's also the stage where you introduce them to a lot of your your educational material with the hope that a subset of them end up going out there and and adopting your product, launching it in production. That's where, hopefully, some of the revenue cycles for the business starts.

That's where the business flywheel that you talk from a developer perspective starts to get triggered. And also, eventually, you want a subset of them to be a strong advocate. The most successful developer teams that I have seen are the ones where people talking about their product out there are not just the the people who are paid to do that, like the DevRel people, but folks from the from the community. So if you have a very vocal and and supportive ecosystem of developers out there who are using your platform and and they are strong advocates out there and you are able to share their stories and testimonials, that really goes a long way. And then we have been using GenAI actually across this this life cycle, going all the way from connecting with third party developers to engaging with them, through content and other ways, helping them adopt, supporting them in that adoption journey, and then eventually advocating and helping them move in in that direction.

So we'll we'll start by covering some examples for that. The first one is how are we using GenAI in support function, which directly contributes to to adoption and and conversion. So at Qualcomm, we have this large global commercial operate operations team that has existed, and their responsibility is to support developers, customers, partners when they get stuck with something. And and there is a email alias that gets shared. If you are a developer above a certain threshold from a from a size and and a support perspective, you can just email this alias, and and someone will get in touch with you and respond.

Now, historically, a lot of this support came manually. But but recently, what has happened is, all of these inquiries that come in get triaged, completely through automation, and then we are also using generative AI to to generate responses that get sent back to the team. There there is also, like, language translation that happens as part of it. Responses are personalized. How has this helped?

This has helped in terms of the accuracy of the response that goes out, the time in which you are able to get back to the customer. So the customer or the developer is not waiting for several days or even weeks, which also impacts the the funnel of conversion. If they reach out and don't hear for a week, they probably find something else. But if you are proactive and able to get them get back to them sooner, it it just changes, the the whole dynamics. The other areas, where we have really used GenAI is I'll start with the first one where when I started at Qualcomm, a lot of our documentation, and this is not not even GenAI, was behind login screens because a lot of documentation was meant for large partners, and and we didn't want it to to just go out and be out there in public.

Transforming Documentation with AI Tools

So we worked on, okay, like, what documentation can be put in front of a login screen, where if you've come to developer.qualcomm.com, we don't want you to see a login screen as the first thing. What is publicly available documentation that exists? That resulted in almost 50,000 pieces of external documentation suddenly being available. Now one way of letting developers navigate that is through okay, like, you from doc document, you're find what you need, and then hopefully, are successful and you move forward. The other way for that is, can that be changed into a a search box where a developer tells you what they are looking for?

And it's not just a search. Like, then a chatbot can respond by going through all of those pieces of documentation and giving you exactly the response and then doing a two way conversation on top of it. So we we build a an internal version of it. We tried it, iterated on it. There were a ton of issues to deal with from just the accuracy to people's asking all kinds of things.

Like, they may ask things that are not meant for developer support, but trying to get information that exists somewhere in your company. Right? Like, so all of that privacy and security risks you have to take care of. But then eventually, we got to a point where we were able to put this out in beta. So right now, 30,000 external customers and all of internal Qualcomm users are able to use this interface to really get going faster.

Now how does this help from a DevRel perspective? Reimagine your, documentation website. Can that go from static pages to a smart chat interface where a developer comes and asks exactly what they're looking for, and you are able to give them that response quickly. Extended to the next stage, can the content be in that shape as well where a developer comes and says, show me a five minute video of what your API does and and if you can include use cases around it. And the video gets generated on the fly.

And then the developer at the next question can ask, okay. Like, this is interesting. Can you can you share a longer tutorial on this, maybe two hours? And then you are able to serve that. And through personalization, you are able to capture the journey.

So the developer is only able to go for an hour and come back later, you are able to do that. So that's that's one area where I think a lot of this work could be extended. Of course, using GenAI as an assistant in authoring documentation, of course, you don't push it out as is. You you take the inputs, you review it, and then then roll it out. But that's another area where where we have been using GenAI.

Using AI for Content Generation and Translation

The third piece is generating content. And before I play some of these videos, I just wanna call out, right, this is not a replacement. Think of some of these things as augmentation. We have always talked about, okay, like, a lot of our developers in country are in countries where English is not the primary language. A lot of our developer I don't have enough developer advocates.

They're ours for them to go and record these videos and do the code samples and go to go at events. Right? So so don't the point here being don't immediately go and replace all of your humanly generated content with AI content, but the point is, like, some of this is actually getting good quality and and and and can help serve those needs that you have for your third party developers at at a much lower cost.

Video voiceover: In this video, we are going to download and install Microsoft Visual Studio, a powerful yet adaptable IDE for your next application.

Video voiceover: First, go to the Microsoft Visual Studio downloads webpage, visualstudio.microsoft.com/downloads. Select the enterprise version. Free trial. The download will begin automatically. Open the setup execution file to see the download in progress.

Uttam Tripathi: One the status in downloading. Movement and the sound number of components because of the setup. To watch the real video, there's movement.

Video voiceover: Select the following three options. Dot net desktop development, desktop development with c plus plus and Windows application development.

Video voiceover: In the individual components tab, add the following components by typing the following into the search bar. C plus plus Universal Windows Platform. Version one four three ARM 64. Also search for MSVC version one four three Versus 2,022. C plus plus ARM build tools latest.

Click install.

Video voiceover: Please note due to the large file size, the installation will take a while. Once the installation is complete, you will be able to access Visual Studio and all of the required components.

an area which: So these are just four AI avatars out of dozens that the team has been experimenting with. And, again, like, we are careful how we are putting it out there. So if you will go to our YouTube channel, you will still still see a lot of like, I think majority or almost all of them is humanly generated content, but this is an area which is of of constant interest. And I think this is one of the places where DevRel is gonna get transformed as we go along. This is another example of a real person.

He recorded this talk in English, but then using lip sync, we were able to translate it into Turkish. So, again, like, that use case of my developers are in different geographies and they need content in their own language, I I think that gets served in a much, much cost effective way.

Video voiceover: Important step in a Qualcomm AI certification program journey. We hope you enjoyed immersing yourself in the world of Qualcomm AI and adding new skills to your portfolio. Qualcomm So

the last video: even the lip movement had had changed with the language. And and this finally is, again, like, a combination of AI avatars, offering instant translations in 100 plus languages. Right? So I'll just play this one. This is the last video.

Video voiceover: And software stack for rapid prototyping and proof of concept evaluation.

we always go: The the point being for some of this is we always go to our leadership, and I I'm I'm in the process of doing that, asking OpEx and budget for next year. We have a number of line items. And and and in the current environment, everything is being looked at with a lot of scrutiny. So if there are certain areas where you can take money from and invest in where you really Like, for us, doing in person workshops is critical, but because that really helps people get over the hump. Or for us being able to to be out there more and and spending time with developers is important, we all have to do trade off because flat budget is the is the new good news.

Like, if if if if I hear flat budget in in any of my conversations with people across the company, I've I get a sense, okay, this is the best case scenario. But you can't operate with flat budget if you're trying to grow the reach of your community and you're trying to do more with developers. So these are some use cases how we are using, GenAI. Again, it's not an exhaustive list, but I think and I'll be very curious over the next couple of days, hear from some of you as well, how are you how are you using AI in in DevRel itself, right, in in in moving forward. Couple of other areas, analyzing and supporting support forums, automating responses over there.

AI's Role in Understanding Developer Needs

This, again, like, I think has been done in the past as well at at scale. And then finally, capturing the pulse of developer community. And and we haven't finished this journey yet, but one of the things I've always found powerful is if Monday morning, you can serve a dashboard to your exec leadership that really talk gives a high level pulse of what your developer community is. What are the five pain points? What are the open standing issues?

And where the community is doing well. Now in order to do that regularly at scale for millions of developers, you can't do it humanly. And and therefore, the power of AI is is is great because then some of those issues can be translated into feature requests and be served back to the product teams and help iterate quickly, like, helping improve that that satisfaction score for your for your community. So that's all from my side. As I said, like, from a Qualcomm perspective, it's a new journey that we are getting into.

You might have not heard about Qualcomm and open source in the same sentence three, four years ago, but here we are, also very committed about, working on opens open source standards, making sure that our APIs are robust, our is our documentation is high quality and accessible. Our SDKs are are really best in class and something that you all can trust, and also starting this new developer go to market motion. And and and and still early days, so we wanna be upfront about that, but this is a critical area where we will continue to invest and move forward. That's all from my side. I think I'm not sure how we are doing on time, but we're gonna be around and happy to to talk to some of you if you have questions on this or anything else.

That's all. Thank you.