AILA: IIT Delhi Develops AI Agent That Can Perform Lab Work And Automate Complex Experiments
Researchers at the Indian Institute of Technology (IIT) Delhi have developed an innovative artificial intelligence (AI) agent named AILA (Artificially Intelligent Lab Assistant) that can autonomously conduct scientific experiments in laboratory environments. This groundbreaking technology aims to enhance research efficiency and accuracy by automating complex lab tasks.
Introduction to AILA
Imagine instructing a robot to carry out a complex laboratory experiment, from setting up equipment to analyzing results. AILA is designed to do just that. This AI agent has been developed through collaboration with various institutions in Denmark and Germany, allowing it to utilize different types of laboratory equipment, make decisions during experiments, and provide scientific analysis akin to human researchers.
Efficiency Boost in Research
Indrajeet Mandal, a PhD scholar at IIT Delhi and the first author of the study, noted that AILA significantly accelerates research processes. He stated, “Earlier, it would take an entire day to optimize microscope parameters for high-resolution, noise-free images. Now, the same task is completed in just seven to ten minutes.” Mandal works under the guidance of Professor N M Anoop Krishnan and Professor Nitya Nand Gosvami, who are both involved in the development of AILA.
Collaboration and Development
The project was the result of extensive international collaboration, involving researchers from Aalborg University in Denmark, the Leibniz Institute of Photonic Technology in Germany, and the University of Jena in Germany. This diverse team has brought together expertise from various fields to create a robust AI framework.
What Sets AILA Apart
The research team discovered that while many AI models excel in theoretical tasks, they often struggle in practical laboratory settings that require adaptability and contextual judgment. Mandal likened this to the difference between knowing driving rules from a textbook and navigating busy city traffic.
Safety was another crucial consideration. The researchers observed instances where AI agents deviated from instructions, highlighting the need for robust safeguards to prevent accidents or damage to expensive laboratory equipment.
A Breakthrough in AI Applications
Professor Krishnan emphasized that AILA represents a significant advancement in the application of AI in scientific research. Unlike traditional large language models such as GPT or Claude, which can answer questions based on existing knowledge, AILA connects these models to real-world laboratory experiments. When instructed to perform an experiment, AILA autonomously writes the necessary code, operates scientific instruments, collects data, and analyzes results.
Mastering Complex Instruments
A key component of this research is the Atomic Force Microscope (AFM), a highly sensitive instrument used in materials science. Operating an AFM typically requires years of training, as it involves understanding nanoscale physics and real-time feedback control. Mandal shared that it took him nearly two years to become proficient in using an AFM. However, with AILA, this learning curve is eliminated. A new researcher can simply provide a prompt and run the experiment within minutes.
Building AILA: From Concept to Reality
The development of AILA took nearly a year, primarily due to the need to create a new framework from scratch. Professor Krishnan noted, “This is one of the first efforts of its kind globally. Once the framework was ready, we conducted numerous experiments to test its reliability.” The team designed 100 different experiments, with AILA successfully completing 80 of them, establishing benchmarks for reliability and performance in autonomous experimentation.
Innovative Framework and Originality
To clarify concerns about originality, Mandal explained that AILA is not trained on past experiments. The system relies on extensive documentation, including manuals that span nearly 4,000 pages. AILA reads these manuals to design and execute new experiments based on user prompts, similar to how a human researcher would approach a task.
Real-Time Adaptability
Professor Krishnan highlighted that AILA differs significantly from conventional lab automation systems, which typically follow fixed, pre-programmed steps. Real-world experiments are dynamic, and AILA can adapt strategies in real time, responding to experimental conditions much like a human researcher.
Safety Measures and Human Oversight
Addressing safety concerns, Professor Krishnan noted that multiple safeguards have been integrated into the system to prevent damage to expensive instruments. While AILA provides recommendations, the final decision-making authority remains with human researchers, ensuring a balance between automation and oversight.
Future Prospects and Scalability
The researchers are now focused on improving AILA’s accuracy and developing an indigenous large language model that can be integrated with the AI agent. This will reduce reliance on commercial AI systems and further enhance AILA’s capabilities. The goal is to make AILA accessible to a broader scientific community, democratizing advanced research tools and methodologies.
Conclusion
The development of AILA marks a significant milestone in the integration of AI within scientific research. By automating complex laboratory tasks, AILA not only enhances efficiency but also opens new avenues for innovation in experimental science. As this technology continues to evolve, it has the potential to transform the landscape of research across various scientific disciplines.
Note: This article is based on the latest developments as of December 2025 and reflects the ongoing advancements in AI technology within scientific research.
