CodeComplete is an AI-powered coding assistant and developer productivity suite built for enterprises. Pulse 2.0 interviewed CodeComplete CEO Max Lu to learn more about the company.
Max Lu’s Background
Lu is a co-founder at CodeComplete, where he leads efforts to build enterprise-level AI coding assistants to enhance developer productivity. And Lu said:
“In my role, I’m actively collaborating with industry leaders to identify and tackle the challenges they face in adopting AI tools for enterprises, including topics such as self-hosted and fine-tuned LLMs. Before CodeComplete, I was leading teams at several startups, including Pilot.com and Autolist.”
Formation Of CodeComplete
How did the idea for CodeComplete come together? Lu shared:
“When some of the first AI coding assistants came out, my cofounder, Max, wanted to use it at Meta. Because of the sensitive nature of his work, the leadership team empathically said no. We realized that there was a gap between the most powerful devtools and the needs of enterprise users, so we created CodeComplete to bridge that gap.”
Core Products
What are CodeComplete’s core products and features? Lu explained:
“CodeComplete is an AI-powered coding assistant and developer productivity suite built for the enterprise. We offer products like code completions (similar to GitHub Copilot) and chat interface (similar to ChatGPT), but differ in a few major ways:
Secure infrastructure: We deploy within your infrastructure so that your IP never leaves your security boundaries. This is the best way to protect against data breaches and IP leaks.
Data training integrity: We offer models that are trained on permissively licensed code, so you don’t risk adding copyright or copyleft artifacts into your codebase.
Customized solution: We fine-tune our model directly on your codebase so it incorporates your libraries, coding patterns, and institutional knowledge. The results are more relevant and effective.”
“CodeComplete offers a suite of customized tools to supercharge the entire developer workflow. We can help developers ramp up quickly on unfamiliar parts of the codebase or help you digest large PRs. We can refactor your code to move from one framework or even one language to another. We can help you automatically write documentation or unit tests for your code, so you can spend less time on boilerplate and more time building the next big thing.”
Favorite Memory
What has been Lu’s favorite memory working for the company so far? Lu reflected:
“Right before our launch with our first customer, the team made a significant improvement to the product. When we tested it out with the customer, there was a magical moment when the customer said, ‘Wow, it read my mind!’ I think that was a great example of the team working together and delighting our customers. We’re always looking for those magic moments with our coding tool experience!”
Challenges Faced
Has Lu faced any specific bottlenecks in your sector of work recently? Lu acknowledged:
“Given the nature of our product, we have to overcome plenty of technical difficulties. Improving the performance of the model is a difficult and time-consuming task. Sometimes, a small adjustment or trick can improve performance significantly, but sometimes the model will train for a week with no real gains in the end. We are building something new, and we can only do so by constantly learning, innovating, and listening to feedback. Luckily, developers are eager to work with us, and we have a team of fantastic software engineers who are passionate about our mission.”
Evolution Of CodeComplete’s Technology
How has the company’s technology evolved since launching? Lu noted:
“Since the initial launch, our technology has undergone significant evolution and refinement. On the model side, we continue to train the model on billions of lines of code, and the AI has become more adept at understanding complex coding contexts and delivering more relevant and precise guidance. On the features side, we’ve not only expanded our language and platform support, but also introduced new functionality based on user needs and technological advancements, such as auto-documentation, unit test generation, and chat Q&A.”
“We’re always staying on top of new discoveries and user feedback, and hope to continue to develop a tool that is secure, ethical, and loved by our customers.”
Significant Milestones
What have been some of the company’s most significant milestones? Lu cited:
“There’s been a few – when we completed fundraising, which has allowed us to grow our team and focus on our mission; when we landed our first customer and were really excited to work on the use case they have outlined; when our model performed at a level that was competitive with the top tools in the market… I’m sure we’ll achieve many more in the future!”
Customer Success Stories
After asking Lu about customer success stories, he highlighted:
“We work with our customers to determine the right scope and use cases for the AI models that we deploy. This can take many forms – it’s more than just code completions! For example, one of our customers is looking to migrate millions of lines of COBOL to Java, and we’ve seen that models specifically trained on this task vastly outperform much larger generic models. That means we can dramatically decrease the amount of migration time (on the scale of years) and costs associated. This is a common use case that keeps coming up, especially in the financial industry, and we hope to tackle it one by one.”
Total Addressable Market
What total addressable market (TAM) size is the company pursuing? Lu assessed:
“Currently, we are focused on delivering a tool for enterprise developers; they can be software engineers, data scientists, production engineers, and more. In the long run, we think there’s an opportunity to build tools for all functions in enterprise companies, including non-technical ones. We believe that AI will change how developers interact with code, how knowledge workers interact with content, and how society explores new frontiers. We hope to service all these use cases someday!”
Differentiation From The Competition
What differentiates CodeComplete company from its competition? Lu concluded:
“There are a few things we uniquely focus on:
Security & governance – enterprise companies are our bread and butter, and they care a lot about security.
We deploy our product fully on-prem or in their private cloud, so their data and IP never leave their security boundaries. This is the best way to protect against data leaks.
We offer models that are trained only on permissively licensed code in order to minimize any legal risks.
We provide full visibility into the product so companies can understand how their employees are using the product and how AI output is incorporated into their codebase or docs.
Customization – we utilize both fine-tuning and retrieval-augmented generation techniques in order to best incorporate a company’s institutional knowledge.
Companies are often sitting on vast amounts of proprietary data, and one of the best ways to incorporate them is through fine-tuning. This allows us to help solve company-specific problems and improve developer efficiency.”
