– Researched state of the art models, incorporating and presenting the new found knowledge to the team.– Revamped an LLM experimental evaluation system to enable automation, reducing experiment time and steps by 75% and supporting the execution of 500+ experiments.

– Worked on a research paper that tackles the feasibility of using LLMs for mental health disorder diagnosis. I contributed to the developement of the evaluation testbed, which was used to run more than 1,000+ experiments– Applied prompt engineering strategies and conducted comprehensive analysis of the resulting outputs.

– Gained expertise in understanding client requirements and structuring projects, including designing API endpoints and implementing containerization for seamless deployment.– Wrote various DL projects, one involved fine-tuning a Bi-Lingual sentiment analysis transformer which scored 0.68 F-score competing with trained models twice its size.– Enhanced existing company projects such as including new required entities for a NER project, Improved the performance of detection models for a security related project by 0.3 F-score

– Acquired proficiency in processing and integrating 3+ types of satellite data into deep learning pipelines.– Developed and compared various deep learning models, attaining an MSE of 29 for forecasting future pollution readings.

– Reached 0.95 Dice score in the Carvana image segmentation competition, utilizing U-Net architecture.– Engineered a solution for age prediction from voice recordings, employing mel spectrograms and an ensemble of CNNs, achieving an MSE of 20.– Placed in the top-3 in 7 out of 9 DL/ML competitions.– Mentored 12 junior club members, sharing knowledge through code reviews, debugging support, and guidance on covered theory.