May 2025 – May 2026 · 1 yr · Bengaluru, Karnataka, India · On-site

• I am currently working on a research project aimed at developing a foundation model for Indian monsoon prediction at a seasonal scale. The project focuses on designing deep learning architectures capable of generating 30–60 day lead time forecasts, with the goal of significantly outperforming traditional statistical and physics-based models. I have implemented various models including stacked autoencoders, neural networks, regression and decision trees, and convolutional neural networks. My work also involves analyzing existing research and methodologies in monsoon forecasting and improving upon them by applying state-of-the-art machine learning techniques. As the project progresses, I plan to incorporate transformer architectures and attention-based LSTMs to capture complex temporal dependencies and improve long-range predictive accuracy.

Dec 2024 – May 2026 · 1 yr 5 mos · Bengaluru, Karnataka, India · Remote

• At the company, I contributed to building an automated data visualization system that translates user queries and dataset summaries into visual insights using large language models (LLMs). My work involved leveraging Microsoft’s LIDA framework to enable hands-free visualization generation through natural language inputs. I integrated DSPy to experiment with and improve query routing strategies, resulting in enhanced quality and relevance of visual outputs. Additionally, I worked on a natural language to SQL code generation pipeline using Transformer-based models, enabling seamless translation of user questions into executable SQL queries. This allowed non-technical users to extract insights from data without manual query writing.

Sep 2023 – Present · 2 yrs 8 mos · Bengaluru, Karnataka, India · On-site