Technical Leadership: Building Powerful Solutions with Simplicity and Inclusion

Transcript

Shane Hastie: Good day, folks. This is Shane Hastie for the InfoQ Engineering Culture podcast. Today, I'm sitting down with Bhavani Vangala. Bhavani, welcome. Thanks for taking the time to talk to us today.

Bhavani Vangala: Thank you for having me.

Shane Hastie: So my normal starting point with these conversations is, who's Bhavani?

Introductions [01:05]

Bhavani Vangala:  Sure. I'm a lifelong techie and currently I head Engineering Autonomous and we're a fast-growing Series A health tech company right here in Bay Area. We started about six years ago with the goal to simplify the complex healthcare workflows through cutting edge technology and also it's accessible through a powerful developer-friendly secure platform. My career journey started as a software engineer, writing code, breaking things, fixing them, and also learning things along the way.

Over the years, I've led engineering teams in both large companies and fast moving startups, and today I'm really lucky to have built both a stellar engineering team and also a platform that's genuinely changing the way healthcare tech gets built and delivered. To me, tech leadership isn't just about shipping features or checking boxes. It's about it's about solving customers hardest problems and building it in a way that is effortless to the user and that's what we build powerful technology wrapped in simplicity. That's the difference I want to make.

Shane Hastie: So powerful tech wrapped in simplicity, as an engineer on a team, how do I understand what that simplicity is?

Powerful technology wrapped in simplicity [02:20]

Bhavani Vangala: It should be easy. In healthcare workflows, you can see they basically go through these gazillions of documents, and summarize stuff, and go here to the system, and that system, and pull up information, and then they talk... For example, I'm just taking this health scenario here. A doctor who wants to just go in to a patient and all that. So they want to talk to a patient about the diagnosis and everything.

They have to go to multiple systems, pull information through, and summarize all that, and get. And on top of that, they also have to give what the solution is, what's the diagnostics and everything. So for them to go over a lot of all these workflows by themself, and then, finally, getting to a point where they need to just focus on just treating the patient, and not worry about all the other things. So the solution of the product that you create should be very easy. So just, they give a patient name, it just does everything for you in the background, and just comes up with that simple information for you to do it.

So building things, or building solutions of a product like that, is what I want to make.

Shane Hastie: So as the engineer on a team, how do I influence this, versus the product manager who's doing the design, for instance?

Bhavani Vangala: So I think we have to see what the goal is. It's not about convincing anybody to do it this way. I think all of us are working towards the same goal. So whether it is product, whether it is UI, UX, the design, or whether it's engineering, I think all of us have to come together and align on that simple goal. And then, I think everything else falls into place of how we want, and how we want it to be designed.

And it all starts right there. So all the group that's responsible for putting the product have to come together, align their thoughts, and then get this going.

Shane Hastie: More and more in today's environments, one of those members of that cross-functional team is a generative AI robot. How do we work with them?

Is using generative AI in engineering overkill? [04:31]

Bhavani Vangala: Generative AI robots, as of now, we have not gone that far yet, so we just work with people right now. Just because there is a technology, I don't think we just have to work with that, and we are forced to work with that. I don't believe in that. I always go by, what is our customer's problem at hand? What is the efficient solution, and the right solution that needs to be in place? That's what we go for. And, right now, I don't think that robotic AI is mature enough for us to go use that and build anything. So, personally, I don't believe in that.

Shane Hastie: Okay. But as an engineer today, I'm being asked to use the Copilot tools and so forth, and we're seeing significant productivity improvements, coupled with, interestingly, significant increases in defect levels. How do we balance those two? And how do we use these tools to the most effectiveness?

Bhavani Vangala: Yes, of course, the engineering leaders are asking the team to use the Copilot and do things. And I have to tell you, there are a lot of engineers who don't want to use the Copilot. And there's a balance that we have to do. Where does it help you? They just say, "Okay, it increases the productivity of the engineers". But we do have to say, "Does it? Or is it creating more issues than what you had before?" So there's a balance that we have to do.

I think, in terms of using Copilot as a companion, like a partner, there is still ways to go there. We should think about the background things that's happening with the engineers using the Copilot. You're asking it to complete a dot, a parenthesis, or things like that. And it's going through that big, massive AI model to do this. And underlying all the energy stuff that is getting spent for this, just for completing your sentence, or autocompleting your code and everything. So in my mind, I think that's overkill.

But then, there is a real use of wherever you want to use in terms of the error stuff. And it would also be used where it can be integrated with the developer's ID, where it would basically go into real-time information, pull up that real-time information, and show you that this is where the errors are, these are the different libraries, and you're making a mistake here. And it has to evolve to that level. So because the models today are with old information, they are trained up to a certain point.

And in this open source world, you basically see things evolving on daily basis, or even hourly basis. And for you to use Copilot, and you to be using the libraries, the misinformation that it will have, in terms of the old libraries and everything, not everything current, it's not going to be useful. So you're going to make mistakes there. There's going to be defects if a developer or an engineer does not crosscheck things, then it's going to be a problem in production.

So use it in a way that don't rely on it fully is what I would say, based on where things are today. And always verify it. As much as we want to say human verification is not needed, and it's going to do and all that, but, as always, you should have verification process that is covering all the things that you need in that code and all that. So do that verifications diligently before you put things into production. Just because something is giving you some suggestions, it's giving you the autocompletion based on the information that it knows, don't just blindly believe in it. And as I said, for it to go to the next level as a partner, it needs to integrate with the current live information about the libraries, and then pull the data from that to be even more better with defects, less defects and things like that.

Shane Hastie: So the latest version of these models are getting better and better at inference and reasoning. How's that helping us, or is it?

AI reasoning and inference [08:38]

Bhavani Vangala: Ok, let me go into the scientific evolution and then we can touch on the AI part of it. So in the 17th or the 18th century, when Isaac Newton published his work at that time, very few scientists or thinkers asked him to reason it out, right? By 20th century, when Einstein shared his findings, obviously the bar was higher, he faced more scrutiny, more questions and he had to reason that out. And now look at scientists in the 21st century, it's just not enough to share a theory, you have to defend it, debate it, and reason with everyone in your field. So that's the arc of scientific evolution towards and deeper reasoning and accountability. Now compare that with AI.

Right now AI arguably is still in the caveman on phase. We have all used Generally AI and at some point or the other with more follow-ups to your prompts, we've gotten responses that is completely out of context. So it is like it answered quickly but didn't understand the question. So pattern matching got us this far it's impressive and to me, AI is as transformative to technology as semiconductors, Internet and Cloud, but the critical next step is reasoning. Because the issue right now is reliability and in terms of consumer apps, that's annoying, but in enterprise systems, that's a deal breaker. 

So, in my mind we've gone through like these three phases right now and the phase one is, well, this model is fast and phase two is like, wait, is this even right? And now we are entering phase three, which is the reasoning and trust. And the companies that are already seeing serious success with Gen AI is they have a cognitive layer on top, which is where the strategy happens, right? So the real winners are those that are using the AI wisely who have built this cognitive layer on top? The next revolutionary step for AI isn't just better output, it's better inference, it's the explanation, justification, and the self-awareness. We need AI doing peer reviews of other AI and human in the loop feedback to close the reasoning gap and also systems that can think through a response, not just fit one out, right? Because we don't need fast answers anymore, we need answers we can trust. And just like how science evolved through reasoning, so must AI too.

Shane Hastie: So with that reasoning, observability becomes an issue, doesn't it? How do we track that reasoning?

Observability challenges [11:22]

Bhavani Vangala: Absolutely. Observability becomes the key in terms of the process that takes place end to end from prompt response. It's the explanation, justification and the self-awareness. That's something that I touched in my previous response. And how do we know these responses are right? What was the steps that followed to arrive at that response? How can we justify or trace back to the response? So these things become critical and I would say there are three major areas that AI has to improve on.

One is the self-awareness and knowing when to answer, when to stop and instead ask more questions. I can give you some specific examples in the initial versions of Gen AI, you ask a question like who's John? Most Gen AI engines will immediately answer with whatever knowledge it was trained on. It took three generations of versions for it to now ask, give me more context, right? Even today, you may go ask any leading Gen AI engine what is eclipse? It'll immediately rattle off about solar and lunar eclipse and every detail that it knows about it. The AI engines do not stop to ask, hey, which groups are you talking about? Is it Celestial or IDE or something else? At this point, Gen AI can finish your sentence but can't finish your thought.

The next is the AI specificity. Do not bring the world's largest AI model to answer a simple question like in a small domain, that AI will only be corrupted because it with its vast information that it has and the more you know, the more you hallucinate. And that's even personal with humans. You can ask a simple question to a three-year old and a 30 year old and something like what does a Ferrari ride through Napa Valley at sunset feel like? So most three-year old will quickly be able to answer like I don't know and a 30 year old who's not been to Ferrari or the Napa Valley, he or she would just be able to answer that's a cool rich lifestyle based on the information that they've heard, right? So the more you know, the more you hallucinate. That is a problem with large AI models and practice. And practice peer review is something that we touched before.

Shane Hastie: What are the team and organization culture implications? How does using a generative AI inference and reasoning model as a partner in my programming work, how does that change my old culture, or does it?

Culture implications of AI adoption [13:53]

Bhavani Vangala: It's really two parts to it. One is the model, is it part of the team using it, or is it part of the product? So I think you're asking about the first one, right?

Team, basically, using it for their day-to-day work. It's not that there are some engineers who would want to use the GenAI tools today for their day-to-day work, and they go and cross-reference it. But there are some who just don't want to use it, too. So it's not like we don't force the engineers to use this in their daily job. It's really up to whether they want to use it or not. But there is always a balance. I'm worried about the creativity, the thinking that a human gets towards a solution or the product that we build. So I am worried of the part that it shouldn't affect that. If you want to go cross-check something using the tool, then use that for cross-verification, or you just need to know something.

But when we design a solution, when you implement a product, it's your thinking. It's that organic thinking that comes out from the engineers. That's what we want. So there is a balance, where each individual basically decides whether they want to use this or not. And if it's helping them in the normal routine, mundane tasks or something, they say, "Hey, this is a documentation". They can, of course, use in the documentation part of it. They can give some things, and then it'll give you a documentation, at least initial documentation quickly, and then you can review it, and you can modify it.

So there are certain ways where it helps the productivity, and certain things where we should not. So, basically, we don't say, "Yes". We don't say, "Use it". We don't say, "Don't use it". It's really up to the engineer how they want to use it. So in my mind, as long as you stay to the goal, as long as you achieve whatever we want to achieve, and how we do it efficiently, and in the process, the engineer should be learning things. And should be sharpening his brain, and his skills, and everything. But in the process, shouldn't be losing the creativity and the original thinking that he or she had, as long as you can make a balance there.

Shane Hastie: So shifting track a tiny bit, and moving away from the AI tools, but digging deeper into that culture, what does good engineering culture look like?

Creating good engineering culture [16:25]

Bhavani Vangala: Creating a vibrant and inclusive culture is key, where the work environment is free from stress, allowing everyone to truly enjoy what they do. And this sparks the motivation, curiosity, and drive in everyone. So leaders must work towards this transformative that will make the individuals find happiness and fulfillment in their work.

So how do you do that? First, ensure the work done by engineers always aligns with company goals. They must genuinely enjoy being part of the efforts that contributes to company's growth and goals. And then create an environment where feedback isn't scary, where no one feels odd for speaking up. Leaders need to not only listen, but act. If someone feels stuck, ignored, or out of sync, that's a red flag. And empower the team, letting your engineering team run with their ideas, and make their own calls, can build up the enthusiasm and the happiness at work. It's like saying, "We trust you've got this". And that kind of vibe makes a huge difference. And be there for them, provide the guidance they need, listen to them. They should feel heard and valued. Foster a culture of feedback. Take immediate actions to rectify them. It's all about nurturing and uplifting and welcoming atmosphere at work.

And, also, ensuring everyone feels respected and integral to the team. So recognition and appreciation, giving a shout-out to your team for their awesome work. Also, make sure career advancements automatically happen, based on their smart performance, without them approaching the managers. And, as managers, provide feedback for improvements as needed.

Shane Hastie: That sounds, dare I be skeptical and say that sounds utopian. As leaders, what do we need to do to enable that?

Leadership practices for empowering teams [18:08]

Bhavani Vangala: So that's what I said. You basically have to keep that vibrant, inclusive culture. That's, basically, the key. And the work environment is free from stress, allowing everyone to relish what they do. So for that, what we have to do is, they must genuinely enjoy being part of the efforts that contributes to a company's growth and goals. So you always have to make them included. As leaders, we have to tell them, basically, the goals.

It should always align, whatever each engineer is working for is always aligning with the goal so they feel happy. When somebody works towards a goal and think, "Okay, I'm contributing to this goal, and this is what is getting this company here, that's when I feel satisfied about the work that I do". So you always have to make sure that the work they do is aligning with the goal. And you have to talk to the engineers of why it aligns with your long-term vision and all that.

And, also, empower the team, letting your engineering team run with their ideas, make their own calls. And that is what brings them the enthusiasm and happiness at work. So we always say, "Okay, we trust you've got this". And that kind of vibe makes a huge difference to the engineers. They feel very confident. They're saying, "Okay, they're trusting this on me".

That, basically, is empowering the team to do things. And then always be there for them and provide the guidance they need, listen to them, and they also should feel heard and valued. And if they provide any feedback, take that seriously. Make sure that it's getting resolved. And then go back to them and talk to them, and say, "Hey, last time, you told me this, is it getting addressed? Do you feel okay?" So, continuously, as leaders, we have to make that effort. The organization is not just going to run by itself. You just say something, "This is what we have to deliver”. And no matter what, we have to deliver this. It doesn't work like that. There needs to be empathy and compassion in the leadership role too. And I think the engineers need to feel empowered and you have to be for them. And, also, there is a recognition and appreciation that needs to happen too. That's when I feel good about working in a place.

Shane Hastie: There's a lot there, but we're also coming out, I would hope we're coming out, of a period when there have been massive layoffs and disruption in our industry. And a lot of that, I want to say a lot of the trust has been damaged. How do we rebuild that trust?

Rebuilding trust after industry layoffs [20:43]

Bhavani Vangala: Yes. When we hire people, it has to be very cautious about, "Do we really need the hires? What do we have upfront? What is our next six months? What is our next one year? Or what is our vision? And how much do we need?" And I think we need to be stringent in terms of hiring, in terms of, do we need it or not? Those decisions have to come upfront. And then not like you hire people, and lot of things in the layoff I've seen, a lot of companies just hire, and then, immediately after a month or so, they're let go. So these things, if they're planned well, I don't think that cases would have occurred. So in terms of hiring, I think you do have to evaluate what are our needs, and then hire based on that.

Of course, there's always the performance is also evaluated and all that. So if we follow this more diligently, at least I don't think you can just gain trust in the matter of a few months or so. But I think, as company, we have to see our hiring strategy and all that from the beginning, and then plan towards having only the necessary, what do we need now? And go forward with that. So at least in a few years or in a few months, they see a change that is happening. And maybe they say, "Okay, this company is doing diligent hiring in terms of whatever are their needs.

So when I work for that company, I know I'm not going to get laid off or fired just like that". And at the same time, it's both ways. It should be employee first. All companies should be employee first. And the company should be doing for employees in the same way employees also for the company. And you have that sync between them. And that's when I think the trust establishes.

Shane Hastie: You're a woman leader in a technology organization. It's sad that, still, that's fairly unusual.

Women in technology leadership [22:46]

Bhavani Vangala: I know. It's sad, even at this day and age, we're talking about this topic. So women in engineering, it's different and not different. So, first of all, I think we should see it in this perspective to begin with, engineering systems do not care whether it is a man, woman, or anyone else that is architecting, developing, and/or maintaining the engineering system. And neither does the business or investors who bring money, or who they are, how they look, and all that. But does this mean engineering as a field is the same for women as it is for others? No.

I think from the very basics, it is different. I'm saying majority, I'm not going for the minority here, but majority, they do see women bear a large share in the family and caregiving responsibilities. So that will restrict us from spending too much hours at work. There's a myth about this. If we spend less hours at work, does it actually make us less contributors? Definitely not. This is where the smart work comes in. Well, there might be one person that can overachieve along both smart work, putting hours of hard work, as well as there's room to be successful, focus on working smartly. And I think every woman, they need to see what strength they have. And as women, there's ability to multitask, and we also have different perspectives that we can bring to. So they have to look at those things, the perspectives to the products we build, how we lead teams, et cetera.

So, in summary, what I'm trying to say is, just know your strengths and use them to succeed. I don't think we can complain about things we don't have control over, society, culture, et cetera, everything.

Unfortunately, we are just still talking about it. I think those take longer time to fix. So what is in your control, as each of us succeed? Just make sure that you know your strengths, and use them to succeed. And as each of us succeed, I think that is when we can bring this positive change in the long term. So in the short term, do not get that keep you away from the success. So don't worry about all the other things, just focus on your strengths. Things are going to happen. See how you have to move forward, and just march towards that goal.

Shane Hastie: As a man in a team, how do I be a good ally?

Creating inclusive team environments [25:12]

Bhavani Vangala: I think everybody should actively embrace and celebrate the diversity within the teams. It's not about the gender. Each one has their own special qualities and their own strengths. As a team, we have to see what is that strength? And how can we use that to make things better, or to deliver things better? So that's really what we have to see. As leaders, all achievements, promotions, everything, are linked to merit and performance, so leaving no room for doubt why someone was promoted. So we have to foster this transparent environment. So the execs need to walk the talk by setting, clearly, equality goals, making sure leaders are really held accountable and hitting those marks. I think a strong leadership is also key to creating this workplace where everyone feels welcomed and valued.

So as men or women, we all have to go up together, and see how we can work well together. And where are each one's strengths? How is each one being recognized? They all should be the common rules. It shouldn't be based on any gender over here. What is work? Work is basically for you to come, you achieve, and it's purely your achievements and performance is what matters. That should be the only keys, or that should be only the marks they're rated against, not any gender or anything like that.

Shane Hastie: So if I'm a parent today, why would I encourage my daughter to go into a technology field?

Encouraging women to enter technology fields [26:55]

Bhavani Vangala: Yes. So we see that technology has evolved over the years. Everything in the world is going to be shaped by technology, and it'll be fast-growing industry for a long time to come. So we do want more women to be represented in this tech industry to create that balance. So I think it's imperative that, be it any field, and they have to learn technology. That's where the world is heading towards, and we have to keep up with the technology. So I'm not just saying women in software industry, women to run their own businesses. So not just coding or anything, it's anything you want to do these days, it's all about you need to know how to use technology. So I think more and more women should come into this industry. And the tech industry, they should create that balance. So it's very imperative for every woman out there to have enough technology expertise for anything they're passionate about, as the future work will always be associated with it.

Shane Hastie: Bhavani, we've meandered a lot, and some really important and interesting concepts in there. If people want to continue the conversation, where can they find you?

Bhavani Vangala: LinkedIn is the best place for me. That's where they can contact me.

Shane Hastie: And we'll make sure your LinkedIn is in the show notes. Thank you so much for taking the time to talk to us today.

Bhavani Vangala: And thank you. Thank you so much, Shane.

Mentioned:

• Bhavani Vangala on LinkedIn