India AI Summit: AI meets biology to power the future of biomanufacturing

Under the BioE3 Policy, the Department of Biotechnology (DBT) and DBT-BIRAC are driving the BioAI initiative—integrating AI with biology to tackle complex research challenges and accelerate innovation across healthcare, biomaterials and agriculture. With AI infrastructure support via a strategic MoU with the IndiaAI Mission, the session positioned BioAI as a key enabler for drug discovery, genomics and precision medicine, imaging and diagnostics, strain engineering for industrial-scale production of biomaterials and bioproducts, and agricultural innovation.

AI is beginning to do something biology has always demanded but rarely enabled at scale: learn from complex systems, design interventions, and interpret results in a loop alongside humans. In biomanufacturing—where outcomes depend on dynamic biological processes, multi-layered scales (from proteins and cells to organisms and ecosystems), and strict constraints around safety and reproducibility—that shift is not incremental. It changes what is possible.

This was the core theme of the “AI meets Biology to power the future of biomanufacturing” session at the India AI Impact Summit 2026, featuring Dr Anand Deshpande (Persistent Systems), Dr Anurag Agrawal (Ashoka University), Dr Aravind Penmatsa (IISc Bangalore), Dr Ashish M Gaikwad (Praj Industries), Dr Debasisa Mohanty (BRIC-NII), Dr Lipi Thukral (IGIB), Dr Madhura Vipra (Medvolt), Dr Manju Tanwar (Organic Recycling Systems), Dr Rajesh S. Gokhale (DBT / DG BRIC / Chairman BIRAC), and Prof. Santanu Chaudhury (School of Advanced Computing, Ashoka University).

From biology’s complexity to AI’s new role

The session opened by grounding the audience in what makes biology different from many other AI domains: biological functions play out across vastly different scales—molecular, protein, cellular, organism, and ecosystem—and they evolve through dynamic pathways. That complexity is precisely why AI’s role is now being reframed. For the first time, systems are being created that can learn, design and interpret alongside humans—detecting signals that are too subtle for traditional observation and connecting concepts across different biological and scientific domains through multimodal AI.

In that framing, AI becomes both microscope and telescope: a tool to see finer details that were previously invisible, and to connect distant signals that were previously unlinked.

BioAI as national infrastructure: ecosystems, data and scale

A key emphasis was that BioAI is not only a research agenda but an ecosystem agenda. Under the BioE3 policy umbrella, DBT and DBT-BIRAC are attempting to bring academia, startups and industry into tighter collaboration so biomanufacturing outcomes can be accelerated across the national ecosystem.

The COVID period was referenced as a scale proof point for biomanufacturing: vaccine development and production moved at unprecedented speed, showing what is possible when capacity is mobilised with urgency and coordination.

The session also highlighted population-scale genomics as a foundational asset. It referenced that 10,000 healthy genomes have been identified and made available through an independent biological data centre, and indicated a major next step: expanding this effort to 1,000,000 genomes, described as a programme expected to be announced soon. The intent is to create rich phenotype and disease models for research, next-generation therapeutics and precision medicine.

Two tracks: next-generation therapeutics and sustainable biofactories

The discussion was framed around two practical biomanufacturing arcs.

The first focused on AI powering biomanufacturing for next-generation therapeutics—where the promise is to discover and design molecules faster, improve their success rates, and shorten the path from initial idea to experimental validation.

The second focused on AI-driven biofactories for sustainable materials. The context was the chemical industry’s dependence on petrochemical precursors and the sheer scale of molecules produced—cited as 60,000 to 70,000 molecules. Over the next 20–25 years, the session argued, many material categories will undergo major transitions: biodegradable polymers, bio-based chemicals, sustainable materials and carbon capture technologies. BioAI was positioned as a tool to make that transition more precise, scalable and economically viable—especially when combined with India’s agricultural biodiversity.

Design in silicon, validate in the lab

Across both tracks, a repeated theme was the emerging “design-build-test-learn” loop. AI can help researchers model and refine candidates in silicon before committing to costly wet-lab iterations. This includes virtual labs, molecule design workflows, and AI-assisted therapeutic design that gives experimentalists better starting points rather than replacing experiments altogether.

One example discussed was antibody design—where AI enables designing antibodies in silicon, testing them in labs, and building pipelines that can reduce dependence on animal-derived antibodies. The point was framed as both scientific acceleration and a sustainability gain, while maintaining that experiments remain essential.

From narrow ML to generative BioAI loops—grounded in reality

From an industry lens, the session traced a shift from manual data curation and early automation to generative AI systems that can optimise multiple parameters at once. Enzyme engineering was used as an example where efficacy, toxicity, immunogenicity and manufacturability all matter simultaneously—requiring multi-parametric optimisation rather than single-metric tuning.

The discussion also acknowledged a key limitation: generative AI can hallucinate. The remedy proposed was grounding models through experimental feedback—design, build, test, share, and feed results back into the system so the models update rapidly. The claim was that such loops can reduce iteration timelines by up to 50% in certain use cases, while creating platforms that can extend to other enzymes, therapeutics, antibodies and related modalities.

From shape prediction to shape reasoning—and the compute question

Another thread connected protein structure prediction to the next frontier: reasoning. If structure is “shape,” the next step is using shape to reason about binding, complexes, targets and movement. That implies deeper models and higher compute—because biomanufacturing needs not just predictions but realistic simulations of biological dynamics that are hard to observe experimentally.

In this view, AI will increasingly complement first-principles approaches by modelling what is intractable in the lab, while experiments validate and correct what AI proposes.

Human roles in an automated future: curated data and high-throughput validation

Even as automation expands—robotic synthesis, robotic experimentation, and closed-loop optimisation—the session stressed that humans remain central in non-negotiable areas. The first is curated datasets: biology may not always require “large” datasets in the language-model sense, but it requires highly specific, accurate, metadata-rich datasets that anchor models in experimental truth. Proprietary, gold-standard curated data was positioned as a core value layer.

The second is validation platforms: training the next generation is not only about building models, but about building high-throughput systems that can test in-silico candidates rapidly and generate feedback that improves models over time.

Training for BioAI: breaking silos across disciplines

The closing focus shifted to skills and institutional change. A core challenge identified was that Indian universities have historically operated in silos—chemistry, biology and computer science separated in practice. But AI-for-biology is pulling those walls down, drawing computer science students into biology problems because the algorithms now have visible real-world impact.

The priority areas described were improving model accuracy for experimental requirements, strengthening experimental throughput and validation, and building interdisciplinary talent that can operate across the full loop—from data and computation to lab work and manufacturing constraints.

AI can compress discovery cycles

BioAI was framed as a national capability shift, not a niche research trend. If India can build the enabling layers—curated datasets, high-performance compute, cross-sector collaboration, and interdisciplinary talent—AI can compress discovery cycles, increase precision, and help biomanufacturing scale across therapeutics and sustainable materials. The promise is ambitious: living factories, data-driven design, and faster translation from biology to industry. The constraint is equally clear: without strong data, validation, and ecosystem alignment, the promise will not convert into production reality.