Insights      Artificial Intelligence      Investing in Georgian R&D to Drive Startup Growth

Investing in Georgian R&D to Drive Startup Growth

While Georgian’s key focus is investing capital in promising startups,  we also invest in our own research and development (R&D) team to accelerate growth in our portfolio companies.

When Georgian was founded in 2008, we focused on identifying key disruptive trends—at the time, it was applied analytics—that were creating new waves of software companies. Our goal was to identify these trends early and invest in companies that were in a position to take advantage of these innovations. 

Since then, we have worked closely with our portfolio companies to help them adopt and accelerate those trends, fueling their growth along with our own. As part of that mission, our R&D team plays an important role in supporting portfolio companies in a process that can be time-consuming and expensive. 

We sat down with Parinaz Sobhani, our Head of Applied Research and ML to discuss why we invest in R&D and how this benefits our portfolio and CoLab companies as well as the broader AI ecosystem.

How Georgian Solutions Can Help Companies Beyond Investment 

About five years ago, we began focusing on applied AI. We invested in technical R&D hires at Georgian to conduct AI research based on developments in the academic community and at places like Google.

We grew our R&D team initially to develop our platform to identify, score and rank companies and markets for our investment pipeline. As our team grew our own capabilities, we realized there was an opportunity to work with our companies on applied AI problems that could generate high return on investment (ROI) and value for the business, so we started collaborating on their projects as well. 

Last year, we realized the potential to use data and technology to help our portfolio companies with lead generation, qualification and intelligence. 

We’d like to believe that’s just the start. Our current roadmap focuses on building technology to help our companies in mergers and acquisitions (M&A), sales automation and other operational areas. The Georgian team uses insights derived from the proprietary data in our platform to advise portfolio companies on strategy. For example, when given a M&A or go-to-market strategy, the platform can recommend qualified targets for acquisition or leads for sales. We see an opportunity to offer more value to founders and CEOs by taking a digital-first, data-driven approach to addressing the challenges of scaling a business. 

How Can the Applied Research Team’s Unique Backgrounds Benefit the Companies They Work With?

Across the firm, but especially on the R&D team, many of us come from technology and technical academic backgrounds, and have already worked at startups. So, we can empathize with founders and what they’re going through. 

Each member of my team focuses on one or two areas of applied research that they are able to leverage to help our companies overcome specific business challenges. We then share the research back with all our companies.  We identify problems that are common across our companies, work on addressing them and accumulate the expertise needed to solve them. 

How Does the Georgian R&D Team Kick-Off its Collaborations?

Often, we’ll start working with a company through a deep dive workshop on Applied AI, Conversational AI or Trust. We start by understanding the company’s business problems, product roadmap, and current maturity, then tailor the workshop based on their needs and requirements while ensuring actionable outcomes. 

For example, for a company barely a year into adopting AI, we help them identify opportunities to add AI into their core offerings, prioritizing and ranking those opportunities based on value.

For a company more advanced in its AI journey, we narrow in on one or a couple of specific AI opportunities. We walk through an AI project canvas to determine actionable steps to execute that opportunity.

Coming out of these workshops, we have a much better idea of where we can add the most value working with each company. The next steps might include: 

  • Hiring a team to deliver on the opportunities identified
  • Arranging a hackathon to quickly deliver on a high-value product feature 
  • Helping the company to adopt our open-source software toolkits to accelerate their AI and natural language processing (NLP) product development
  • Scoping out a longer-term engagement to solve a specific business problem

Whatever opportunities we identify, we align with the company’s leadership on the next steps before proceeding. Our Guide to Building ML Teams and our Guide to Building Conversational AI Teams both include more in-depth advice on organizational structure and hiring.

How Do We Work with Companies to Help Them Differentiate and Build a Data Moat?

With more off-the-shelf ML, NLP, and speech processing tools now available, companies can, over a weekend almost, build a rudimentary system that works pretty well. This is a challenge for those who have spent the last few years building a moat around their in-house ML/NLP technologies since AI technologies have been democratized over the past few years and are much more accessible. This means that technology itself is no longer a sustainable differentiator; now, there is an opportunity to build a moat around data and technology. 

However, we see that there’s a continuing need for verticalized data or proprietary data. Data labels especially are still incredibly valuable. So, if you have labeled data, or, even better, a mechanism to efficiently label data, that’s going to set you apart. Our team has built a toolkit to make the data labeling process more efficient. 

We’re also building tools to help our companies take pre-trained models and then rapidly adapt them to a particular use case using a small data set representing the problem you’re solving. We see huge opportunities for companies to maintain differentiation and actually lead the market this way.

Tell Us What an Engagement with the Georgian R&D Team Looks Like. 

One of our portfolio companies, DISCO, is in the legal discovery space with a mature AI team and several ML solutions already in production. We started with an AI workshop to identify their most important business challenge and what could move the needle.

We focused on onboarding new customers faster to address the cold start problem and upgrading their AI systems to be more accurate and data-efficient. Working with DISCO’s data scientists, we used our applied research team’s knowledge of NLP representation learning, transformers and cross-customer data to help DISCO adopt the BERT language model and create automatic out-of-the-box text classifiers. As a result of the work, DISCO was able to create new higher quality models in a quarter of the time and DISCO’s customers can accomplish their work faster and more accurately. 

Having better, more cost-efficient and data-efficient models gives DISCO a much clearer return ROI and value proposition to differentiate them in the marketplace and attract new customers. Check out our podcast to learn how DISCO is cracking eDiscovery with the help of AI

How can companies start an engagement with the Georgian R&D team? 

Georgian’s R&D efforts are not only improving our own efficiency, but also benefiting our portfolio and CoLab companies. We are happy to help with tools and advice for R&D and all other aspects of our companies so we can grow together. If you’re working at one of Georgian’s companies, reach out directly to me.

We also welcome any and all requests and ideas from our CoLab participants, to help us discover your areas of concern. To reach out, get to know each other, and see how we can engage, email our CoLab Product Owner Conor Ross.

Read more like this

Why Georgian is Investing in SurrealDB

The proliferation of unstructured data has, in our view, made building modern…

Redefining Legal Impact with the Team at Darrow

When we think about legal tech software, we think about value add…

Testing LLMs for Trust and Safety

We all get a few chuckles when autocorrect gets something wrong, but…