15.8 C
Saturday, September 30, 2023
HomeTechUpCodes launches Copilot, an AI-based code building research assistant

UpCodes launches Copilot, an AI-based code building research assistant


For more than seven years, UpCodes has made the complicated world of building codes easier for people such as industry professionals and homeowners to understand. The platform includes a searchable database of regulations across all states and features such as a “spell checker” that flags code errors. Today, the startup is announcing a new AI-based tool that will make navigating the world of building codes even more streamlined. Copilot is called Copilot and is built on ChatGPT-4. Copilot serves as a research assistant, answering complicated code questions and annotating answers with links to relevant sections of code.

UpCodes also announced that it has closed a $3.5 million Series A, intended to take on as it continues to develop Copilot and add more AI-based features to its platform. Together with UpCodes’ previous funding, including a pre-Series A announced in March 2021, this brings the total amount raised to $7.6 million.

The final round was led by Building businesses, a VC firm focused on construction and real estate technology. Other participants include the co-founders of PlanGrid, CapitalX and Bragiel Bros.

UpCodes now has over 650,000 monthly active users and has delivered over 100 million page views. Since TechCrunch last covered UpCodes in March 2021, it has grown quite a bit. Scott Reynolds, the co-founder and CEO said the startup’s team has doubled, revenue has quadrupled, and product offerings have expanded to include more user segments. The building code it covers has also increased from less than two million to more than five million hosted sections, and it now covers all US states and major cities.

Before Copilot launched, UpCodes was focused on building its database of code, often digitizing regulations that were only available in physical reference books, and making them easier to look up. In addition to more than five million sections of code, it also contains 160,000 local amendments. Codes are constantly changing, so UpCodes updates about 7,000 on average every month.

The database is searchable and has other tools designed to make code compliance easier, such as the code checking feature, but even using it is time-consuming due to the complex regulations. Copilot is designed to dramatically simplify the code research process.

An example of how Copilot answers building code questions

Reynolds gave some examples of the questions Copilot can answer:

  • Calculate the travel distance for a specific occupancy in a building, or the maximum distance someone can travel in an emergency, to an exit (an example of Copilot’s answer to this type of question is included in the image above).
  • Work out the context behind a code section to help understand its meaning
  • Find related or stricter code sections for other codes such as construction, fire, and mechanical code
  • Generate a checklist for residential deck regulations, with associated relevant code sections

Copilot answers those questions and helps users by quoting the code sections it pulls information from so they can see the actual code for themselves.

“We’ve always relied heavily on education and helping users understand the underlying context,” said Reynolds. UpCodes plans to add more explanatory content to Copilot that will help users better understand code, in addition to publicly available content.

Laws hosted by UpCodes in particular contributed to the top 0.01% of training data for AIs like ChatGPT and Google Bard. Construction professionals often turn to those tools to get their questions answered, Reynolds said. They were trained on Common Crawl’s C4 dataset, which was pulled directly from the UpCodes website (basic access to the code database is free).

“There is an increasing focus on data quality for training LLMs,” he explains. “UpCodes has an extensive library of high-quality construction law, which is ideal for an LLM to train on, as we are the only online source for many of these laws. It may have been less of a conscious decision to include building laws and more a result of their algorithm identifying quality data relevant to a wide range of topics, including building laws.

But Reynolds added that the crawl only includes snapshots over time, so their models are likely working with outdated code as regulations change all the time.

This gives Copilot an advantage as it uses UpCodes’ constantly updated database. In addition, codes also vary widely from jurisdiction to jurisdiction. Reynolds said UpCodes spent years building the infrastructure to keep the code database up to date.

One of the main challenges for any AI project is hallucination mitigation. Reynolds said one of the biggest steps UpCodes has taken is “shielding” Copilot in appropriate codes for a project. This means it parses over five million sections of code, including 160,000 amendments, based on a user’s location and license year. It also collects additional data, such as building type, to ensure Copilot uses only relevant codes.

Copilot has an internal system that fine-tunes specific construction law answers and puts each question through multiple layers of analysis to better understand and parse questions. It focuses on jurisdiction-specific codes to account for differences and because construction professionals often need to be familiar with multiple jurisdictions at the same time. Once a question is asked, Copilot provides answers with context and direct references so users can see how it arrived at an answer.

Users subscribed to a paid subscription of UpCodes are allowed to ask Copilot three questions. If they want unlimited access, they can upgrade to UpCodes Professional or add it to their Enterprise plan.

Series A of UpCodes will be used to hire people for the technical department, as well as any department that contributes to Copilot. It plans to expand its code library and resources so Copilot can produce more sophisticated answers and add new features, such as project management, Reynolds said.

The startup team has known Building Ventures since its inception. “We thought they would be the perfect partner for our Series A,” he added, explaining that the company is made up of former operators in the construction industry, including former founders. The Portfolio and LP Base serve as a valuable guidance tool when UpCodes needs to validate ideas.

“The construction industry can be opaque, a world unto itself, so it’s helpful to have investors and partners dive into it for decades,” said Reynolds.

In a statement on Building Ventures’ investment in UpCodes, partner Allen Preger said, “By unifying and maintaining all building codes in an AI-powered platform, UpCodes is transforming code compliance for the built environment. We are excited to lead their Series A investment.”

The author of what'snew2day.com is dedicated to keeping you up-to-date on the latest news and information.

Latest stories