Harshini Kumarasubramanian

I am a Specialist Programmer at Infosys Ltd with expertise in machine learning and foundation model development. My work focuses on creating efficient AI solutions that solve real-world problems.

Education

• Master of Science in Computer Science - [Your University Name], [Your City, State] (YYYY - YYYY)
• Bachelor of Technology in Information Technology - [Your College Name], [Your City, State] (YYYY - YYYY)

Skills

Research Publications

• "Title of Your ArXiv Paper 1"

  Authors: Harshini Kumarasubramanian, et al.

  Date: YYYY-MM-DD

  🔗 View on ArXiv

  [Brief abstract or key contribution of the paper]

• "Title of Your ArXiv Paper 2"

  Authors: Harshini Kumarasubramanian, et al.

  Date: YYYY-MM-DD

  🔗 View on ArXiv

  [Brief abstract or key contribution of the paper]

• "Title of Your ArXiv Paper 3"

  Authors: Harshini Kumarasubramanian, et al.

  Date: YYYY-MM-DD

  🔗 View on ArXiv

  [Brief abstract or key contribution of the paper]

• "Title of Your ArXiv Paper 4"

  Authors: Harshini Kumarasubramanian, et al.

  Date: YYYY-MM-DD

  🔗 View on ArXiv

  [Brief abstract or key contribution of the paper]

Research interests

Foundation Model Training

I specialize in developing and fine-tuning large language models for specific domains. My research explores optimization techniques that reduce computational overhead while maintaining model performance. I'm particularly interested in:

• Efficient training methodologies for large-scale models
• Knowledge distillation and model compression
• Domain adaptation for specialized applications
• Interpretability and bias mitigation in foundation models

Statistical Modelling and Slow Science

I believe in rigorous methodologies that prioritize reproducibility and scientific integrity over quick results. My approach includes:

• Bayesian methods for uncertainty quantification
• Robust statistical analysis that goes beyond p-values
• Open science practices and transparent research workflows
• Critical evaluation of machine learning methodologies

Current Projects

• Efficient LLM Fine-tuning: Developing techniques to adapt foundation models to specialized domains with minimal computational resources
• Explainable AI Systems: Building tools that make complex model decisions more transparent and interpretable
• Cross-modal Foundation Models: Exploring the integration of multimodal data in foundation model architectures

My Projects

Working together

If you are a scientist who wants to collaborate on ML/AI research or applications, please reach out to me!