May 2024 – Jul 2024 · 2 mos · Bengaluru · Hybrid

• Developed an automated prompt optimizer for aligning LLM outputs with human preferences, using automatic instruction generation & selection, focused on classifying ad spaces used for promotion, achieving 87% accuracy.
• Optimized instruction by searching over a pool of 15 instruction candidates proposed by LLM to maximize a score function, evaluation with zero-shot performance of another LLM; use of semantically similar instruction variants.

May 2023 – Jul 2023 · 2 mos · Michigan, United States · Remote

• Engineered and fine-tuned OCR models for accurate recognition of handwritten text and chemical structures. Architected and trained machine learning algorithms for automated assessment. Conducted extensive data preprocessing and augmentation to get significant results.
• Implemented advanced feature extraction techniques to support grading algorithms. Optimized model hyperparameters and performance metrics for superior reliability. Collaborated with academic researchers to align project objectives and deliver regular progress reports.

Jan 2023 – Apr 2023 · 3 mos · Sydney, NSW · Remote

• Aimed at developing advanced machine learning models for forecasting the stock prices within the fnancial market
• Leveraged a diverse range of deep learning models, including Convolutional Neural Networks(CNN), ResNet-101 & ResNet-50
• XGBoost & Transformers to comprehensively analyze & optimize the datasets for accurate prediction outcomes
• Implemented strategies to efectively handle the class imbalances using SMOTE, undersampling, class weights
• Ensures accurate analysis & prediction; has potential to empower investors & improve decision-making in fnancial market

Jan 2023 – Mar 2023 · 2 mos · New Delhi, Delhi, India · On-site

Mar 2022 – Apr 2022 · 1 mo · New Delhi, Delhi, India