2024 MIT R&D Conference: Startup Exchange Lightning Talks - Delineate

EMILY NIEVES: Hello, everyone. My name is Emily Nieves. I'm co-founder and CEO of Delineate. I am a PhD candidate in biological engineering and CSAIL, and I have about seven years experience building systems pharmacology models at companies like Pfizer and AstraZeneca. My co-founder, Jawad Iqbal, is an expert on digital transformation solutions and AI. And Delineate is a multidisciplinary team. It's growing. We have about 20-plus engineers and researchers dedicated to applying AI to accelerate analysis workflows.

So computational analysis can be really powerful. This can look different in different fields. It can be building AI models. It can be doing more traditional statistical modeling. But it has a really big problem in that the workflow is really scattered. It takes a lot of manual effort to put together structured, clean data sets from very many disparate sources. So we work a lot in the biopharma space. And talking with our customers, it can take teams of scientists months to put together a single data set.

So Delineate's goal is to make that faster using our custom trained models to process all of this unstructured information into insightful assets. So we work a lot with research papers, specifically the plots and graphs, getting that numerical information out. But we also process websites and databases. We use our models to structure this, to clean it, and deliver data sets that are fit for purpose, but also code to generate those models to analyze that data. And we can also generate reports.

So we're building an all-in-one research web tool that starts from search. So AI enhanced search to help you find those documents that have the data that you need. All of that gets put into one place. And then, of course, the meat of our tool, our really core competency, is that extraction of the data. Getting that numerical information out, but combining that with the contextual understanding of where that data came from. And we also have a platform where you can analyze the data inside that same tool using our AI assistant. So the right model with the right data.

So as a case study, we were able to generate a data set for AI model training for a top pharma company. So we processed over 900 research papers. Using our custom computer vision methods, we extracted the numerical information behind all the graphs, combined that with our LLM agents that gave a contextual understanding of how that data was collected, and that was packaged into a data set that was essentially plug and play for their AI model. So we generated 23,000 data points for them.

So what's really different about us versus other solutions you could go to? So really, we are a fit-for-purpose data extraction. Any data set that you can think of, we can help you build. So the inputs to this would be a semantic description of the goal of your analysis and the data criteria. So what should the columns look like? And then the sources get processed using that. So Delineate will extract the relevant information from across the text, tables, and figures, and then put it together in this analysis-ready data set.

So really, we have these fine tuned models for data extraction. That really is another thing that sets us apart. We're all in one place, which simplifies the workflow. Saves time. And we're also 100% auditable. So you can see exactly where every data point came from. Why a line of code was created.

So we've made a lot of traction to date. I'm really proud of what our team has accomplished. So we have data service contracts with two top 10 pharma companies, and we're going to be acknowledged in some of their upcoming publications. We recently became a preferred vendor of data sets for one top 10 pharma company. We're going to speak at the American Society of Clinical Pharmacology and Therapeutics soon. And we received early acceptance to Y Combinator's winter 2025 batch. So that will be starting soon.

And we're really looking to partner, really, with any companies that have needs related to data mining and structuring of complex information. But we have a lot of traction in biopharma, so we would really love to talk to anyone in biopharma as well. Please reach out to us in our email. Or you can find us on the table outside. Thank you.

SPEAKER: Thank you so much, Emily.