The integration of Large Language Models into the biological sciences has transitioned from theoretical potential to measurable wet-lab performance. At the vanguard of this shift is CRISPR-GPT, an AI agent capable of managing the complex reasoning required for gene-editing experiments with remarkable precision. Recent case studies demonstrate that these agents can achieve approximately 80 per cent editing efficiency on first attempts by automating guide design and experimental mapping. This evolution is timely, as the global biotechnology market is projected to exceed 3.8 trillion USD by the early 2030s, driven largely by the convergence of generative AI and synthetic biology.
However, the leap from successful laboratory “miracles” to repeatable clinical pathways requires more than just raw compute. As the Asia-Pacific drug discovery market accelerates at a nearly 10 per cent annual growth rate, founders must navigate a fragmented regulatory landscape. In hubs like Singapore and Jakarta, the challenge is twofold: maintaining technical excellence while adhering to stringent data residency laws such as Singapore’s Health Information Bill. This necessitates a “security-first” infrastructure where biological reasoning occurs within digital vaults that protect sensitive patient data and intellectual property.
Here are the reasons why founder-led bottlenecks are becoming silent growth killers for Southeast Asian startups
We speak to Dr Ilya Burkov, Global Head of Healthcare and Life Sciences, Nebius, about the strategic imperatives for biotech leaders in 2026. Beyond the initial allure of AI-driven efficiency, we examine the “expert-level” metrics that truly resonate with regulators—moving past simulations to demonstrate clear clinical lines of sight. From managing the “visiting” model of decentralised research to mitigating the risks of GPU-heavy infrastructure, this discussion provides a 12-to-18-month playbook for scaling healthtech innovations in a region facing a projected shortage of nearly 7 million healthcare workers by the end of the decade.
Your team has published work and case studies on CRISPR-GPT that show researchers hitting about 80 percent editing efficiency on a first attempt in specific wet-lab demonstrations. What parts of that workflow did the AI agent handle end to end, and what still required expert judgement?
The agent functions as a digital co-pilot that manages the intellectually heavy segments of the gene-editing lifecycle. It handles the end-to-end reasoning for selecting the right editing system, designing the genetic guides, and mapping out the experimental plan based on simple instructions.
Expert judgement remains essential for the final validation of the AI’s designs and the physical execution of the experiment, as the tool cannot perform manual lab procedures like cell culture or transfection. Scientists are also vital for navigating unexpected data, troubleshooting complex biological challenges, such as immune responses, and handling unique organisms that fall outside standard models.
Crucially, while the system includes built-in safeguards, human oversight is mandatory to ensure all work meets strict ethical standards and safety regulations. For example, the system is programmed to identify and block unethical requests, such as those involving the modification of human embryos or viral pathogens, by immediately alerting the user and terminating the session, thereby preventing misuse.
What does expert-level results mean in practice for early biotech teams? Which metrics should they report, and which ones can mislead investors and regulators?
For early biotech founders, achieving expert-level results in 2026 means moving beyond basic science and proving that your technology can actually become a product used by doctors. It is no longer enough to show that your technology works in a computer simulation; you must demonstrate a clinical line of sight. This means showing that you have a clear path to turning your discovery into something tangible, like a pill or a cell therapy, for a specific group of patients, such as those with a certain genetic mutation.
To gain credibility in today’s market, founders must focus on two critical pillars. First, the mechanism of action, or the biological how-to manual detailing exactly how treatments interact with the human body to fix a problem; Second, a clear strategy for compressing the traditional 10-to-15-year drug development cycle by demonstrating how AI tools can eliminate specific research bottlenecks.
This is particularly important in the Asia-Pacific region, which is currently one of the world’s fastest-growing drug discovery markets with a projected growth rate of nearly 10% through 2033. Beyond technical expertise, leadership teams must master the legal and regulatory landscape: this means securing composition-of-matter patents to protect the unique chemical blueprint of their drug, while simultaneously developing the strategic agility to navigate the diverse and often complex healthcare regulations across Southeast Asian markets
For a startup in Singapore or Jakarta, what is the minimum viable compute and data stack to run serious biological reasoning safely?
Startups in Singapore and Jakarta need a secure AI setup that combines powerful processors with digital vaults to keep sensitive patient information locked away even during analysis. In Jakarta, this system must keep all information on servers to comply with national law. For startups in Singapore, the priority is ensuring systems meet the strict security and interoperability standards of the Health Information Bill, which mandates that all licensed providers contribute key patient data to the National Electronic Health Record (NEHR) to support coordinated care.
This security-first approach is essential for any startup looking to tap into a regional biotech market that is set to nearly triple in size by the early 2030s. Much of this momentum comes from an urgent, local need for better ways to track and treat infectious diseases.
On the clear horizon for this industry is a “visiting” model of research, seen in initiatives like the AI Structural Biology Network. This collaborative concept allows researchers to solve complex biological puzzles in hours rather than weeks, all while keeping their proprietary science and patient records locked safely. By adopting this decentralised strategy, startups can safely turn data into new medical breakthroughs while remaining fully compliant with the unique privacy laws of both cities.
How should healthtech founders think about vendor risk, pricing risk, and portability when they build on GPU-heavy infrastructure?
In 2026, the choice of a high-powered computing provider as a vital laboratory for medical discovery is a high-stakes legal decision. To mitigate the risk of relying too heavily on one vendor, founders prioritise partners who offer advanced security tools and a shared commitment to protecting data.
Managing financial risks involves treating computing power as a dedicated asset rather than a variable expense, which provides a predictable budget during intensive research phases. For a startup to remain portable, the founder should favour open systems over exclusive technologies to ensure that work can be moved between providers as needed. This strategy preserves research independence and allows models to be deployed wherever data resides as regional regulations shift.
What has to change to move from miracle cases to repeatable clinical pathways, and where does automation make the biggest difference, in design, manufacturing, monitoring, or documentation?
Turning sporadic medical successes into predictable treatments for everyone requires treating biology as a standardised engineering process rather than a mystery. While technology supports every stage of this transition, its impact is most significant in monitoring and documentation, where it handles the data-heavy tasks that usually lead to staff exhaustion. In the manufacturing phase, robotic precision removes the risk of human error in production, ensuring that complex therapies can be scaled reliably.
We must consider the fact that Southeast Asia faces a projected shortage of 6.9 million healthcare workers by 2030. By automating the review of medical reports and the standardisation of data across borders, technology handles the clerical burden that prevents clinicians from working at the top of their license. Ultimately, by letting machines manage the oversight of health data and paperwork, the region can scale advanced care without losing quality.
If you had to give Southeast Asia-focused founders a 12 to 18-month playbook, what would you prioritise first: automating lab and analysis steps, securing compliant compute, building partnerships with hospitals, or generating real-world evidence?
For a healthtech startup to succeed in Southeast Asia, the first eighteen months must be a disciplined transition from infrastructure to evidence. The journey begins with securing compliant compute, as establishing a high-performance foundation is the essential prerequisite for complex biological reasoning and ensures strict adherence to local data residency laws.
Once this foundation is in place, the focus shifts to automating analysis steps by deploying AI co-pilots that can compress research and development cycles, turning what used to take years into a matter of months or even days.
Finally, the founder should prioritise building hospital partnerships to generate real-world evidence. In a region where non-communicable diseases account for 55% of all deaths, claiming 9.5 million lives annually, collaborating with local clinical clusters is a strategic way to validate tools against regional genetic variants and is critical to securing the clinical adoption required for long-term impact.