Anukriti

Anukriti uses Gemini AI to simulate complex real-world scenarios, helping innovators and researchers explore outcomes, detect risks early, and design better solutions before deployment.

Comment

Inspiration

Modern drug development faces a major challenge: the same medicine can produce very different effects across individuals due to genetic differences. Many drugs fail late in clinical trials because these variations are discovered only after expensive testing.

We were inspired by the idea that AI could help researchers explore these differences earlier. If we could simulate how drugs interact with genetically diverse populations before trials begin, researchers could detect risks sooner and design safer and more inclusive treatments.

This led to Anukriti, an AI-powered system that explores pharmacogenomic variations and simulates possible drug responses across different genetic profiles.

What it does

Anukriti is an AI-driven research assistant that helps explore how genetic differences may influence drug responses across populations.

The system allows users to input information such as:

  • Drug name
  • Genetic variants (e.g., CYP450 genes)
  • Population characteristics

Using AI reasoning and pharmacogenomic knowledge, Anukriti generates insights about:

  • Potential metabolic differences
  • Population-specific drug sensitivity
  • Possible adverse reactions
  • Variations in therapeutic effectiveness

Conceptually, the system explores relationships like:

$$ Drug\ Response = f(Genetics, Metabolism, Dosage, Population) $$

By simulating these interactions, Anukriti helps researchers think about potential risks before clinical trials begin.

How we built it

Anukriti combines AI reasoning with biological data interpretation to simulate drug response insights.

The system architecture includes:

Frontend

  • Interactive interface for entering drug and genetic information

Backend

  • Python-based analysis layer
  • Data processing for pharmacogenomic relationships

AI Layer

  • Gemini models used to reason over drug–gene relationships
  • AI-generated explanations and risk insights

Knowledge Sources

  • Public pharmacogenomic datasets
  • Scientific literature
  • curated gene–drug interaction information

The overall workflow:

  1. User inputs drug and genetic context
  2. Backend structures the pharmacogenomic data
  3. Gemini processes the information
  4. The system generates reasoning-based simulation insights

Challenges we ran into

Building a system that connects AI reasoning with biological concepts came with several challenges.

Data complexity

Pharmacogenomics involves many interacting variables including enzymes, metabolic pathways, and population genetics. Structuring this information in a usable format required careful design.

Scientific interpretation

AI can generate insights, but those insights must remain scientifically grounded. Ensuring that outputs were meaningful rather than speculative required prompt design and validation.

Model reasoning limits

While LLMs are strong at reasoning, translating biological relationships into structured simulations required multiple iterations and experiments.

Accomplishments that we're proud of

  • Designing an AI system capable of reasoning about pharmacogenomic relationships
  • Building a working prototype that connects drug data, genetic context, and AI analysis
  • Demonstrating how AI could assist early exploration of drug response diversity
  • Creating a foundation for future research tools in computational pharmacology

Most importantly, we proved that AI can help researchers explore population-level drug response questions in an interactive way.

What we learned

Working on Anukriti taught us several important lessons.

AI can act as a research collaborator

Large language models are capable of assisting in scientific exploration when guided with structured prompts and contextual data.

Biology is deeply interconnected

Drug responses are influenced by multiple genetic and metabolic factors, making simulation-based exploration a promising approach.

AI tools can accelerate early-stage research

While not a replacement for laboratory experiments, AI systems can help researchers generate hypotheses faster.

What's next for Anukriti

The current prototype demonstrates the concept of AI-assisted pharmacogenomic exploration. The next steps include:

  • Expanding integration with genomic datasets
  • Improving drug–gene interaction modeling
  • Building population-scale simulation capabilities
  • Developing an interactive research dashboard
  • Enabling researchers to test multiple drug scenarios simultaneously

Our long-term vision is to create a virtual population simulator for drug development, helping researchers explore treatment safety across diverse genetic groups before clinical trials begin.

Built With

Share this project:

Updates