📄 Curriculum Vitae
💼 Experience
Working on research under Prof. Mahesh Mohan at the Signal Processing and Artificial Intelligence (SPAi) Group, Department of Artificial Intelligence, IIT Kharagpur, with a focus on SAR LCC. My research spans Remote Sensing Foundation Models, Representation Learning, Knowledge Graphs, Transformers, and Complex-Valued Neural Networks for SAR understanding and improved generalization across diverse geolocations. Presented research work at IndoML 2025 and the IndiaAI PreSummit 2026.
Contributed to the development of reliable conversational AI systems by designing robust parsing logic for structured LLM outputs. Focused on aligning model-generated responses with downstream actions and developing multi-step interaction flows to ensure seamless, tool-call based interactions and high consistency across user sessions.
Worked on financial forecasting for client datasets using statistical, probabilistic, and classical time-series models, along with transformer-based architectures (IBM Granite / Moirai). Evaluated model performance across scenarios and contributed to selecting the most suitable approach for client use cases.
Worked on a research project analyzing the cost-efficiency of ISRO’s satellite launches in comparison to the US space sector. Involved in developing an artificial neural network model to impute missing satellite launch cost data, strengthening dataset reliability for time-series and comparative analysis. Contributed to generating analytical insights that supported the study’s exploration of frugal innovation in the space industry.
Worked as part of a team developing intelligent financial web agents to extract real-time bank interest rates, leveraging automation for accurate data acquisition. Involved in integrating Omniparser to streamline data parsing and enhance extraction accuracy and overall system performance.
🎤 Conferences & Presentations
- Selected to present SAR-Foundation, a novel Complex-Valued Vision Transformer for SAR representation learning
- Selected to present 'Tri-CSWANet' at the IndoML 2025 symposium, selected from numerous national submissions