Manasi Acharya

Created ML evaluation workflows used across multimodal projects to support faster experimentation. Reduced 40% of manual effort in experiments by standardizing inference and evaluation pipelines. Improved generative model efficiency from 10.4% to 77.7% by optimizing latent representations to generate more accurate images. Analysed model performance across demographic subgroups to evaluate bias and reliability in predictive outputs. Tech Stack: Python, PyTorch Lightning, ViT, Docker, Slurm, Git, TensorBoard

Reduced manual data preparation by 70%, automating ETL pipelines. Identified process inefficiencies in team workflows and reduced low-value meetings, improving delivery targets by 15%. Tech Stack: Python, Plotly, Seaborn, SAP HANA, Airflow, Kubernetes, Docker

Improved system performance by 25% through SQL query optimizations to handle high-volume transaction processing reliably. Developed a claim risk prediction model and achieved 92% classification accuracy to support risk assessment decisions. Designed real-time KPI dashboards for ticket status and complexity, reducing overdue tasks by 20%. Tech Stack: Python, scikit-learn, PowerBI, Matplotlib, Oracle PL/SQL, Pandas