Radical Reads

Intrepid Labs emerged from stealth in May, announcing $11M in funding. In this feature, Intrepid Labs CEO and Co-founder Christine Allen explores how AI and robotics are transforming the critical field of drug formulation.

In the high-stakes world of pharmaceutical development, discovering a promising molecule is only the beginning. Before any drug reaches clinical trials, it must be carefully formulated into a safe, effective and manufacturable product, whether as a tablet, injection, nasal spray or other delivery format.

Drug formulation is more than mixing ingredients. It’s a multidimensional design challenge involving chemistry, materials science, pharmacokinetics, and manufacturing constraints. If a drug is the passenger, the formulation is the plane: it must deliver its cargo safely, reliably, and under extreme constraints. For something as “simple” as an oral tablet, there are billions of potential combinations of excipients and processing parameters that must be considered.

Traditionally, formulation R&D has relied on legacy know-how, guided trial-and-error, and incremental iteration. This creates a self-limiting cycle where researchers gravitate toward what has worked before, leaving vast unexplored territories in the formulation design space. The result is slower development, limited innovation and high clinical failure rates.

Incomplete and biased training data compound the challenge. Much of the industry’s proprietary formulation data remains locked away, and negative experimental data rarely sees the light of day. This creates a skewed perspective where AI models trained on available data miss the full complexity of drug formulation.

A New Approach: Self-Driving Labs Meet Drug Development

At Intrepid Labs, we have developed a fundamentally different approach. Our AI-driven robotic platform, Valiant, does not rely on massive historical datasets. Instead, it creates a closed-loop system that can start with no data and rapidly explore vast formulation spaces through intelligent experimentation.

Here is how it works: We input the target product profile for a specific formulation, the drug itself, and parameters for the formulation design space. Our proprietary algorithm then selects which formulation to prepare and test first. The automated platform creates and analyzes this initial formulation on a small scale, feeding the results back into the algorithm for retraining.

This iterative process continues through several batches, with each experiment informed by the results of the previous ones. Unlike traditional approaches that test one parameter at a time over months, we can optimize multiple characteristics simultaneously, from tablet size, hardness and stability to drug release profiles and material costs.

Beyond Speed: Reducing Clinical Trial Failures

This approach offers more than just speed. Better formulations could dramatically improve clinical trial success rates. Many promising drugs fail in clinical trials not because the active ingredient lacks efficacy, but because suboptimal formulations prevent the drug from reaching its therapeutic potential.

Our platform can also breathe new life into existing drugs. At Intrepid Labs, we are reformulating drugs that are already off patent or are about to go off patent, creating new formulations that outperform existing market standards. 

The Future of Pharmaceutical Development

We are entering an era where the combination of AI and robotics can unlock formulation possibilities that were previously impossible to explore. The convergence of these technologies represents a new paradigm in pharmaceutical development, in which delivery format, dosing schedule, and patient experience can be intentionally designed from day one rather than constrained by historical precedent.

As we scale our platform and expand our partnerships with pharmaceutical companies of all sizes, we are working toward a future where the vast majority of clinical failures due to suboptimal formulations become a thing of the past. 

Find out more in Intrepid’s announcement and website.