Hire Akshitha B. - Fine-Tuning Engineer

I work on building AI-driven decision systems that translate business objectives into scalable, production-ready machine learning solutions. I lead end-to-end development of predictive models for risk and outcome scoring, combining structured experimentation, feature engineering, and threshold optimization to improve model recall and reliability in high-impact scenarios. I focus heavily on model interpretability and deployment-readiness — implementing explainability techniques and building lightweight inference pipelines to ensure models are transparent, robust, and seamlessly integrated into internal workflows. My work has directly contributed to measurable improvements in operational efficiency and model performance.

* Led a multidisciplinary team building AI models for medical imaging and radiotherapy dose prediction for head and neck cancer patients. * Designed and experimented with diffusion-based and Mamba-inspired architectures (DoseDiff, CT Mamba) for 3D dose prediction on the OpenKBP dataset. * Managed project execution end-to-end: data preprocessing, training workflows, experiment tracking, and performance benchmarking using dose and DVH metrics. * Collaborated closely with faculty and teammates to refine research direction, troubleshoot model failures, and improve clinical relevance of predictions. * Presented results, ablation studies, and visualizations to academic audiences, highlighting trade-offs between model complexity, stability, and accuracy.

developed and optimized supervised machine learning models for both classification and regression tasks, improving validation performance by 12% through advanced feature engineering and systematic hyperparameter tuning. I built end-to-end ML pipelines handling data preprocessing, transformation, validation, and performance evaluation across structured datasets. I worked closely on cross-validation strategies, bias-variance diagnostics, and production-aligned evaluation metrics to ensure model generalization. Additionally, I automated training and inference workflows and assisted in exposing prediction services through API-based endpoints for internal testing and integration.