Crystal-XG

Crystal-XG is an AI-driven platform revolutionizing pharmaceutical drug development. Over 80% of drugs require crystallization to reach patients, yet the process remains costly, inefficient, and largely unchanged for a century. By leveraging a neural network trained on the world's largest crystallization database, we predict the exact conditions needed to grow any crystal, reducing a process that once took months and hundreds of thousands of dollars to minutes. We exist to accelerate the journey from molecule to medicine, making drug development faster, cheaper, and more sustainable for researchers worldwide.

Crystallization is a cornerstone of modern science. Over 80% of pharmaceutical drug products require at least one crystallization step in development, and the process is equally critical across chemistry, materials science, and beyond. Thousands of laboratories worldwide are dedicated exclusively to crystallography. Yet despite its fundamental importance, fewer than 1% of candidate drug molecules ever reach approval. Solubility and crystallization failures remain among the leading reasons promising compounds never reach patients. The problem is not a lack of scientific effort. It is a lack of innovation in method. For over a century, crystal synthesis has relied on trial-and-error experimentation, guided by theoretical intuition but driven largely by guesswork. A single compound can require hundreds or thousands of individual experiments before viable crystal conditions are identified, consuming months of researcher time, hundreds of thousands of dollars, and significant quantities of solvents, reagents, and energy that are ultimately discarded. The environmental and economic toll is enormous, and the field has seen remarkably little fundamental change. Crystal-XG was founded to change that. We leverage machine learning, specifically gradient boosting models (XGBoost), to predict the optimal conditions required to crystallize any client-provided molecule. At the core of our platform is the world's largest crystallization database: over 10,000 unique crystallographic experiments, capturing chemical components, synthetic procedures, and key reaction parameters. This database is ten times larger than anything else available globally, and it is the foundation upon which our predictive model operates. A client inputs their molecule; our model outputs the exact solvent system and synthetic technique needed, reducing a process that once took months to minutes, at a fraction of the cost. Our model has been validated through the successful crystallization of diverse inorganic macromolecules, achieving 90% accuracy. This work has earned Crystal-XG first-place finishes at the 2025 Palmetto Pitch Competition, 2025 Launchpad Liftoff Competition, and 2025 College of Science Catalyst Competition, as well as the Clemson BuildFund microgrant, totaling over $15,000 in non-dilutive funding. We have filed a provisional patent through Kim, Lahey & Killough Law Firm and are actively collaborating with research groups worldwide to publish our findings. Most recently, Crystal-XG became the youngest company ever backed by the South Carolina Research Authority (SCRA), a milestone that reflects both the credibility and urgency of our work. The need for Crystal-XG extends well beyond pharmaceuticals. Energy storage, catalysis, and advanced materials research all depend on crystallography to understand molecular interactions, predict material behavior, and design compounds with desired functionalities. In every one of these fields, the same inefficiencies exist, and the same opportunity for transformation awaits. We envision a future where predictive crystallization replaces guesswork, where the environmental and economic burden of traditional methods is dramatically reduced, and where novel materials with vast scientific and industrial applications are discovered faster than ever before. Crystal-XG is not just making crystallization more efficient. We are changing how science moves molecules from the lab to the world.