Anand Kulkarni
Appearances
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
And then we take a look at what steps we need to carry out to improve those capabilities in a place that's going to have the biggest impact for our customers based on what our customers tell us that they think they want. and on the metrics that we have in our own forecast for how this works.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
When we talk about requirements capture, turning a code base into requirements, we have a core metric of accuracy that we look at, and we have an internal mandate that we have to be the most accurate capture tool in the market. Out of a candidate list of activities we can carry out, we can look at pipeline investments, we can look at ways to speed up the import process so that we can do more runs.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
We can also look at ways that we can present more information back to the user about potential issues with their import. We rank each of those things, we take input from customers, and then we decide, based on that information, what is the biggest priority that we should carry out that's going to have the biggest impact on this one metric that matters for the business.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
But you can see that it's all informed by data and it's informed by customer input.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
We've got a lot of great people here at Cribotics. My CTO, Charles, has been with the company almost since its first year. He's really one of the technical visionaries behind what we build in code to spec. He, like me, has this enduring curiosity and passion for how to do work at the bleeding edge of LLMs. And he, like me, spends a lot of time thinking about what that looks like.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
We've got a number of strong folks up and down the staff. In terms of what we look for, I look at a handful of things to help me assess if folks are really strong, but belief in the strategy, the vision, willingness to push hard. And of course, I also tell people, we're signing up for a home run here.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
You should be prepared for what that will take in terms of investment of your own effort and energy. And you should be driven and motivated by that as an upside and reward. And that combination, when paired with raw intelligence, ends up making for a really powerful team.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
I'm really happy to say that the folks that we have here are some of the folks that I've been very proud to work with and alongside. You really want to have great co-workers in the trenches there with you. And that's something we've been able to do.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
So scalability is this question that as soon as you solve it at the level that you are aiming for, your own ambitions take you up to the level where you need to start thinking about it again in the same way. I don't think there's any universe where you can avoid working on this problem unless you are not willing to grow or not able to grow.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
We have been in a couple of periods in our growth where we have seen that we were catching up on the ability to support the scale that we were at because our demand suddenly spiked as we were more successful with customers. Good problem to have, but still a real and massive problem that you've got to deal with or you lose credibility.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
We generally like to build for and test for substantially more scale than we currently have. But of course, there have been a couple of times before we learned that lesson in practice where we had to scramble to meet demand. And in those cases, you want to make sure you have a team that's willing to step up and be in the trenches with you.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
I am really excited by the idea that we have been able to make such a dent in a problem that was so intransigent for so many organizations. We have managed in a universe that is drowning with AI software engineering tools to make a product that is somewhat unique in terms of its ability to successfully improve, optimize, and scale the software development lifecycle.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
Really, we position ourselves as the AI product manager that runs alongside the AI software engineers that are out there, but that's something that nobody else has been able to really do in the same way we have. I'm most excited by the idea and most proud of the idea that we saw the trend coming. We made the right decision to jump on board with it.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
We've been really successful in being able to deliver something really good for our customers. I think at the same time, that means we've been able to build something we're proud of as an organization and something that our own team is excited to work on and deliver.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
I'm optimistic that we will continue to stay in front on this specific set of challenges and we'll do our best to keep delivering in the way we have for our customers.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
Before we went all the way in on modernization, there was this persistent debate on how deep we wanted to go in supporting Greenfield application creation versus requirements modernization in Brownfield. We built quite a bit of a product line that we ultimately killed, which was a visual drag and drop low code editor. It was cool.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
It ran on top of your code base and it ran on top of your requirements. But ultimately, our customers did not care about that. Not in the least. They said, there's a lot of low-code tools out there. We're excited about Crowdbotics, not because it's low-code, but because you are able to generate requirements really quickly. And I think there was this idea, we had this sunk cost fallacy.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
So scalability is this question that as soon as you solve it at the level that you are aiming for, your own ambitions take you up to the level where you need to start thinking about it again the same way. I don't think there's any universe where you can avoid working on this problem unless you are not willing to grow or not able to grow.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
We had spent so much time investing into this particular line of business, and we were excited by what we thought the technology potential was, that we did not listen to our customers quickly enough. What we should have done was made the difficult decision early to mix that product, to drive focus inside the organization and shut down the studio product line sooner. That was a mistake.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
We should have done that. And I think with experience, now we know focus is the key that drives you at scale in ways that are much harder if you decide to try and do everything.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
We're all in on code to spec right now. When we look at the future of how this thing's expanding, today we're supporting a lot of mainframe code, COBOL, FORCRAN, some stuff like PS1, and some specialized systems, and then lots of cloud code, .NET, Java, PHP, even modern stuff like Django Rails node, all of which is being plugged into Gradbotics Processing.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
We're excited to continue to expand those frameworks that we support and excited to support a broader set of types of customers and types of activities within the SDLC. Our bread and butter is still modernization. It's the part that most companies end up coming to us to support.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
But we're also seeing emerging interest in related areas like maintenance, migration, all of which basically run through the same workflows modernization does. Look, I think the category here where I'd like us to end up is as the category leader and winner on how teams reverse engineer code bases.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
To make sure that happens, we've got to keep investing in the core quality of our model to make sure we are delivering the best and highest quality results. I'm also looking forward to seeing where we can also go beyond just systems like modernization, migration, and maintenance. We're asked to look at things quite a bit that are not quite code, but almost code if you sprint at it.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
And I think there's potential here for LLMs. and for Crowdbotics to look at those things as well. I'm talking about things like EHR migrations or CRM migrations, things that are really painful processes that enterprises and organizations have to do with human beings. We know that AI agents can solve this stuff. We're just not working on that today. So...
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
I'll look forward to seeing us able to support problems like that over time, especially as we get deeper into the market. We'll be able to look at everything else that we might want to generate requirements for turning into a crowdbotics object and being attached into our system.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
I was inspired by my own parents more than anything else. My mom was a teacher. My dad worked in energy innovation for the state of California. They put a good work ethic in me. They also put a lot of curiosity in me as a kid. And I just remember them taking me to the library all the time. I would come home with arms books and that gave me a real drive to keep working to finish the next book.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
spent a lot of time thinking about ideas as a kid. So I think that's been a piece that has definitely motivated me. Look, they say our parents are our first teachers, and I think they're also the ones who give you the way that you think about the world. I owe a lot to them for that kind of thinking and that kind of decision-making.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
We have been in a couple of periods in our growth where we have seen that we were catching up on the ability to support the scale that we were at because our demand suddenly spiked as we were more successful with customers. Good problem to have, but still a real and massive problem. I'm Anand Kulkarni, CEO and founder of Crowdbottom.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
There are a lot of ways to run an organization, a lot of ways to build a product and launch it. I think the thing that I want people to think about when they are launching their own product, who are you solving it for? How do you make their lives better? And how do you make sure that you are relentlessly laser focused on solving that person's problems?
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
That's the part where if you're able to do that, you will never have a problem as an entrepreneur. If you are really solving somebody else's problem and solving it, you're in business. And that's basically what it comes down to. So for that person sitting next to me on a plane, I'd say, who's your customer? Let's go find them, show it to them. And you tell us or have them tell us what they think.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
Your own passion as an entrepreneur lets you see the world as it's going to be. But you convincing a customer is what turns it into a business. And that's the part that matters when you're launching a new product, a new company.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
We are a modernization company. We use AI systems to turn code bases into human-readable natural language specifications. What that means is we are writing requirements automatically at scale from code bases. You've got a code base, we can reverse engineer that into requirements. Once you've got those requirements, you can flip that back into more code.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
We are helping the world deal with this massive problem of updating, fixing, and modernizing all this legacy code. This is something that system integrators, consultants have been helping entities with for a long time. They do an okay job, but there's just too much code.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
What we do in our core insight is that in the LLM era, human written requirements give you this common currency between what machines can understand and what human beings can understand. Crowdbotics exists to support modernization efforts by providing in natural language, human-readable requirements that are also machine-readable, that allow you to go quite a bit faster. So that's what we do.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
Today, AI-based software engineering is a very crowded field. A lot of these folks maybe don't appreciate the complexity and difficulty of this problem. When you look under the hood, you find out it's just a thin wrapper on OpenAI or Anthropic. We've actually been working on this specific problem for a very long time.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
In 2016, I was looking at early literature about not large language models, but the precursor technology known as recurrent neural networks. I remember reading this great paper by Andre Carpathy, who was at Stanford at the time. It was called The Unreasonable Effectiveness of Recurrent Neural Networks.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
And he showed that with a pretty small program, you could train on a very large amount of data and get something produced that looked almost, but not quite like code. And I looked at this and I said, you know what, this is the future. As this technology improves, this is going to be the way that everyone ends up building software.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
So we set out to build a company that would automate the process of software creation using AMP. We took the approach that we were going to have to build our own training sets and models from scratch. We set out to build essentially a data collection, the ability to capture information about how software was being written.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
And the goal here was to understand how requirements, the things that we use to describe how we build software mapped into libraries so that we can understand the relationships between how people described code that they wanted to build. and then what libraries, technologies, and solutions they wanted to use.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
It turns out the best way to get that kind of data is specifically by running your own development marketplace. And so we ran an open development marketplace for many years. It took us about five years to get all the data we needed. The deal we made was that if customers needed software built, they could come to us. We would match their needs up against development teams who could do the work.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
Customers would keep the code that they built, but we would keep the metadata. The association between requirements and what libraries were being used to satisfy those requirements. And that let us build out this really unique, interesting data set.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
And of course, that was the core of how we ended up building this classical machine learning system that would map requirements into code and code into requirements. Brownfield is the overwhelmingly biggest and harder problem to solve, and it's mostly neglected by industry. We realized that was really the core audience for this.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
When we talk about the MVP, we're really talking about two generations of this company, right? Yeah. Era one was building to collect data and helping people build Greenfield software so that we could collect data that we needed. That era was the first five years of the business. We built the whole stack in Python directly with Django on the back, At the time, it was a view on the front.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
The MVP itself, really V0 of that tool. We built that thing in a couple of weeks for the barest of bare bones version that would just transact and let you minimally place orders, match them up and capture the information we needed. Eventually, we flipped a bunch of the core tech into the modernization engine.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
The modernization engine itself, we built over the course of 12 months to get that core machinery built. And it emerged from the pieces of what we had before being reconfigured in. So our core LLMs stayed the same. The core requirements factor changed from being one where we would capture that data on the outside to one where we would make that data available for customer use.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
We've always built software in-house. Django is our framework of choice. Nowadays, we use React. The ultimate machinery, of course, under the hood, we switched long ago from using PyTorch and our own methodology to being able to run on top of Azure and Azure OpenAI as our main AI backend.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
I'm a technical founder. Even though as CEO and entrepreneur, you don't have to be the technologist, because I happen to be a technologist, I have opinions about the system. We built the early version of the platform on Django. I had an early choice, I remember, between building on Django and building serverless, which was brand new at the time as a methodology.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
But first I tried serverless, and I spent a week... Wasting my time trying to teach myself serverless while building the MVP. I remember thinking, what a dumb decision. At the time, I was just so disappointed. I was like, I can't make this work. I can't believe it.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
But then I realized repeatedly how happy I was that I had simply chosen Django because I knew the system well and because Django is easy to build quickly. Of course, now we do things that are architecturally far more complex than that. But that simple choice we made early on let us go very quickly in a lot of places very early.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
Probably the most difficult decision we made to roll back technically was building on Vue. We built in Vue a whole bunch of front end. And today, very little of our front end is in Vue. We ended up changing it all to React. And that's one that was painful to walk back from, no question. Look, all code is temporary. That's the way to think about it.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
You become very comfortable with driving the right technical decisions and learning to challenge that debt wherever you're able to.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
So if you look at what Credbotics does at our core, we're doing three things with customers. We're doing requirements capture, meaning you show up for the code base. I want to make requirements that describe that code base accurately and completely. The second is we want to make it easier for customers to analyze, extend and modify, reimagine those requirements.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
How do I talk to my code base or talk to my requirements about this code base and say what needs to change? What is working? Where in my code can I change X? And then we want to help customers do requirements-driven development.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
How do I take a set of requirements that I've got, modify those using intelligent agents inside Cloudbotics and inside my AI software engineering tool, GitHub Copilot, which works very well as a combined effort. Okay, if you go through those three things, first, you've got to have a working baseline flow, which is the thing I talked about that took us nine to 12 months to build.
Code Story
S10 Bonus: Anand Kulkarni, Crowdbotics
It gets you all the way from starting with a code base over to I am now building software and modernizing code using this end-to-end system. You've got to do it in a way that gets better over time. So when we look at our roadmap, we are looking at ways that we can assess systematically how good we are at each one of those activities.