◦ Designed and implemented functional chatbots in Python using Hugging Face API and quantized LLMs (Mistral, DeepSeek, Qwen, LLaMA), delivered via 4–5 documented Google Colab notebooks. ◦ Developed a retrieval-augmented generation (RAG) chatbot with web-scraped content (BeautifulSoup), FAISS indexing, and SentenceTransformers to provide context-aware responses. ◦ Authored over 1000 lines of original code, including adaptations for classification outputs from generative LLMs, optimized for limited Colab GPU resources. ◦ Led training runs and analysis for emoji-inclusive sentiment classification models (NLTK, VADER, TextBlob+BART, BERT, DeepSeek, Mistral, Qwen) on 200 labeled samples (emoji-inclusive vs. exclusive), generating detailed performance metrics and confusion matrices. ◦ Co-authored accepted papers at ICTIS 2025 (New York) and NBEA 2025 (upcoming) on emoji-aware AI and AI R&D ROI assessment; presented an abstract at NJBDA 2025 (William Paterson University, NJ). (Mar 2025 - present)

Processed 500 GB of GMRT telescope’s raw data into pulsar candidates, manually labeled over 170,000 pulsar candidates. ◦ Performed elaborate data extraction to obtain 2D images and 1D arrays from pfd source files. ◦ Optimized plots for a deep learning model that improved classification accuracy by weighting features from both data types. ◦ Created a complex deep learning model that integrated features from both 2D and 1D data, achieving enhanced accuracies through the incorporation of customized GANs and loss functions. ◦ Demonstrated the effectiveness of deep learning models in astrophysics, contributing to interdisciplinary advances. ◦ Administered an Ubuntu server, installing custom software like sigproc-4.7, PRESTO, and TEMPO. Streamlined data processing with Python and automated tasks using shell scripting, significantly boosting efficiency. (Jan 2023 - Apr 2024)

◦ Assisted Dr. Vivek Kumar Singh as a Teaching Assistant for ECS102 : Inroduction to programming course with 375 students. ◦ Conducted lab sessions to guide students in mastering fundamental C programming concepts, encompassing data types, variables, assignment statements, loops, and functions. (Mar 2023 - Jul 2023)

◦ Engineered a cutting-edge Convolutional Vision Transformer (CvT) architecture for precise plant leaf disease classification. ◦ Attained state-of-the-art outcomes on an agricultural dataset, surpassing performance benchmarks set by individual CNNs and ViTs. ◦ Demonstrated adaptability through meticulous pretraining and fine-tuning exploration, showcasing the model’s versatility for diverse downstream tasks. (Dec 2022 - Jan 2023)

Analyzed historical sunspot time series data using various forecasting methods, including statistical techniques like AR, MA, ARMA, and ARIMA models. ◦ Explored advanced machine learning approaches such as linear regression, random forest, and XGBoost, as well as neural network architectures like LSTM and Prophet, to improve forecast accuracy. ◦ Developed predictive models for sunspot count and positioning, accompanied by butterfly plots for visualising forecasted patterns. (Mar 2022 - Jun 2022)