When I joined the Department of Statistics at Florida State University in 2007, I thought I’d spend my whole career in academia. Over 15 years, I moved from assistant to full professor, taught, advised students, and ran a lab.
My journey didn’t start in statistics, though. I studied chemistry (BS, MS), computer science (MS), bioinformatics (PhD), then trained in statistics during my postdoc.
During my PhD (with Dr. Jie Liang at UIC) and postdoc (with Dr. Jun Liu at Harvard), I worked on protein structure modeling. In the last year of my postdoc, I shifted gears and started exploring biological text mining.
At FSU, I continued working on protein structure modeling and text mining, but also jumped into genomics data analysis. It wasn’t easy, but it taught me to connect ideas and data across fields.
We participated in several BioCreative challenges biological information extraction (IE) over the years. In 2017, we joined BioCreative VI using traditional machine learning while others used deep learning. We didn’t perform as well, but I told my students afterward: “We’ll learn deep learning and go all in.”
That summer, we took Coursera courses together and did every assignment. It was a turning point.
In 2021, our team joined BioCreative VII and performed strongly across multiple tracks. Later that year came the LitCoin NLP Challenge by NIH and NASA, testing knowledge graph construction — an area I had worked on for years. We put everyone in the lab on the project and won 1st place.
That win changed everything. I saw how fine-tuned LLMs could achieve near-human performance in biological IE. The breakthrough was here, and I knew many real-world applications were about to become possible.
So in the summer of 2022, I took a leave of absence from FSU and started working full-time on Insilicom, a company I had quietly founded back in 2012. When ChatGPT launched later that year, it didn’t surprise me.
Since then, our team has grown to over 20 people. We’ve won the BioCreative VIII knowledge graph construction track (2023) and the BioASQ 2025 challenge for document retrieval and snippet extraction.
Then, earlier this year, I had to make a difficult choice. FSU asked me to either return or resign. By that time, Insilicom had momentum, with people depending on me, technologies ready to scale, and opportunities too big to ignore.
I chose to move forward.
Today, I’m working full time at Insilicom, building AI-powered solutions for biomedical research, drug discovery, and pharmacovigilance. I believe knowledge graphs are the key to creating truly trustworthy AI, and they can transform how we understand biomedical knowledge.
Challenges lie ahead, but even more possibilities. I’m excited for what’s next and will share this journey on LinkedIn and our blog: https://blog.insilicom.com.
#AI #DrugDiscovery #KnowledgeGraph #Pharmacovigilance #Entrepreneurship