In a move signaling a seismic shift in how life sciences organizations conduct research, computational software pioneer Schrödinger has entered into a strategic collaboration and software agreement with pharmaceutical giant Bristol Myers Squibb (BMS). At the heart of this partnership is "Bunsen," Schrödinger’s newly unveiled agentic AI co-scientist. This platform is designed not merely as a data processing tool, but as an active, autonomous execution engine capable of navigating the vast, trillion-scale landscape of modern computational chemistry.
By integrating Bunsen into their research workflows, the two companies aim to solve one of the most persistent bottlenecks in drug development: the bridge between massive computational capacity and the strategic, human-led direction required to turn potential leads into viable therapeutic candidates.
A Chronology of Computational Evolution
Schrödinger, founded in 1990, has spent over three decades laying the groundwork for this moment. The company’s trajectory has mirrored the broader evolution of the industry, shifting from early physics-based simulations to the current era of AI-driven, high-throughput experimentation.
- The Early Years (1990s–2000s): Schrödinger established its reputation by focusing on high-accuracy, physics-based molecular modeling, providing chemists with the tools to simulate atomic interactions.
- The GPU Turning Point (2010s): The emergence of GPU computing transformed the field, allowing Schrödinger to scale methods like Free Energy Perturbation (FEP) to levels previously considered impossible, enabling faster and more accurate predictions of binding affinity.
- The Rise of Integrated Platforms: Through the development of tools like LiveDesign, the company fostered collaborative environments where medicinal and computational chemists could synthesize data from disparate sources.
- The Agentic Era (2025–Present): With the introduction of coding agents like Claude Code in early 2025, the industry witnessed a proof-of-concept for autonomous workflows—systems capable of writing code, debugging, and iterating. Schrödinger has now adapted this paradigm for the laboratory, using Bunsen to automate the complex, multi-step workflows that define modern drug discovery.
The Trillion-Scale Funnel: Redefining Research Capacity
The scale of molecular exploration has undergone an exponential expansion. Five years ago, the ability to enumerate 100 billion compounds around a lead molecule was considered state-of-the-art. Today, that number has surged into the trillions.
As Pat Lorton, Chief Technology Officer and Chief Operating Officer at Schrödinger, explains, this increase in scale is not just about quantity; it is about the precision of the filters applied. "We can easily go into the trillions," Lorton noted. "As the pool has grown, so has the number of properties we can predict, which makes each screen more selective."
Lorton draws a compelling analogy between the current state of drug discovery and the evolution of the aerospace industry. In the mid-20th century, aircraft design relied heavily on physical wind tunnels and manual craftsmanship. Eventually, the aerospace sector moved toward computational fluid dynamics and supercomputing, which allowed for virtual simulation of drag and aerodynamics, drastically reducing the need for empirical testing. Schrödinger is aiming to do for pharmaceuticals what NASA did for aviation: move the "wind tunnel" into the cloud.
Supporting Data and Technical Infrastructure
The partnership arrives at a time when "Big Pharma" is aggressively investing in high-performance computing (HPC) infrastructure. The industry is currently in an arms race to secure the compute power necessary to run complex foundational models.
- The BMS Supercomputing Surge: Bristol Myers Squibb has committed to deploying one of the most powerful AI supercomputers in the pharmaceutical sector, utilizing NVIDIA’s DGX SuperPOD architecture. This infrastructure supports a "Predict First" philosophy, which seeks to integrate AI-generated predictions into every small-molecule program and the vast majority of large-molecule programs.
- Hardware Agnosticism: A critical feature of Bunsen is its flexibility. While it is optimized for high-end environments, the platform is intentionally LLM-agnostic. "The product we’ve built is LLM agnostic," Lorton stated. "It plugs into Claude and works amazing there, but if you install your own open-weight LLM on your own SuperPod, we’ll be able to use that just as well."
Official Perspectives: The Synergy of Human and Machine
The collaboration between Schrödinger and BMS is underpinned by a shared belief that the future of discovery is "agentic"—meaning the AI acts as an extension of the human researcher rather than a replacement for them.
Stephen Johnson, Vice President of Computational Sciences at Bristol Myers Squibb, emphasized that Bunsen allows their scientists to "think differently" about the molecular design space. This sentiment is echoed by Erin Davis, BMS Vice President of Research Business Insights & Technology, who noted that the company is actively building its own foundational models to support large-molecule research.
From Schrödinger’s perspective, the tool is a remedy for the scarcity of highly skilled computational chemists. "Right now, the biggest problem we see in the industry is that there aren’t enough people who are experts," Lorton observed. By automating the arduous tasks of data assembly and workflow management, Bunsen functions as a "postdoc" for busy researchers, allowing them to focus on high-level strategy and decision-making.
Implications: The Future of Drug Discovery
The integration of Bunsen into the BMS workflow carries significant implications for the industry at large.
1. The Death of the "One-Shot" Workflow
Lorton warns against the temptation to treat AI agents as "magic boxes" that can solve a drug discovery project in a single prompt. He likens this to trying to write an entire complex software project in one "one-shot" attempt. Instead, the most successful research teams will be those who steer the AI, utilizing their domain expertise to guide the agent toward specific regions of the molecule or particular hydrogen-bonding pockets.
2. Human-in-the-Loop as a Safeguard
Despite the autonomous nature of Bunsen, Schrödinger has emphasized the importance of human oversight. The platform allows researchers to disable automatic job launches, requiring the AI to present a "thesis" of its proposed workflow for human review. This ensures that even when the agent operates at massive scale, the scientific rationale remains transparent and subject to expert verification.
3. Compounding Accuracy
As workflows grow longer and more complex, the importance of "talkative" software—tools that share data seamlessly—becomes paramount. Lorton argues that if eight different computational products are used in a series and they do not communicate perfectly, errors compound. The Bunsen platform is designed to unify these tools, ensuring that the input from one stage of the funnel provides a clean, validated baseline for the next.
4. Bridging the Talent Gap
Perhaps the most immediate impact of the Bunsen collaboration will be the democratization of expert-level workflows. When a researcher can offload the manual assembly of a calculation campaign to an agent, they regain the time to analyze results and develop new hypotheses. This could effectively "scale" the productivity of existing computational chemistry teams, allowing them to manage more projects with greater efficacy.
As Schrödinger and BMS move forward with this partnership, the industry will be watching closely to see if this agentic approach delivers on its promise of shorter timelines and higher-quality drug candidates. The transition from manual simulation to agent-driven discovery is well underway, and for the pharmaceutical giants, the stakes could not be higher. By combining the massive, trillion-scale power of modern supercomputing with the strategic, iterative guidance of an AI "co-scientist," the path to the next generation of life-saving medicines is becoming clearer, faster, and more precise.
About the Author
