Shobha Dasari

We built three multi-layer perceptrons that can serve as the backbone for a practical prediction tool to measure injury severity in terms of patient mortality, prolonged length-of-stay at hospital, and discharge disposition. Our study population comprised admission encounters of adults aged 18 years and older who were admitted with a primary diagnosis of traumatic injury. We achieved above 0.85 recall (primary performance metric for a screening tool) for all three models on the test sets. We… Show more

We built three multi-layer perceptrons that can serve as the backbone for a practical prediction tool to measure injury severity in terms of patient mortality, prolonged length-of-stay at hospital, and discharge disposition. Our study population comprised admission encounters of adults aged 18 years and older who were admitted with a primary diagnosis of traumatic injury. We achieved above 0.85 recall (primary performance metric for a screening tool) for all three models on the test sets. We also generated Shapley additive explanations (SHAP) values to explore which features had the greatest influence on driving model outcomes and found that factors such as age, subdural hemorrhage, and hip fractures are all positively correlated with patient mortality, prolonged length of stay, and discharge disposition. We foresee these models being used as a screening tool for trauma patients, allowing first responders and clinicians to rapidly triage patients and inform real-time clinical decision-making. Show less

Designed a new product feature set that will help LinkedIn better cater to students' needs and improve the student experience on the platform.

My team and I conducted over 200 survey responses and 50 user interviews to learn about students' professional development needs. We used our findings to create 3 prototypes for new Mentorship, Jobs, and Feed features, and conducted market research to develop a go-to-market strategy for this feature set. Throughout this process, we've also… Show more

Designed a new product feature set that will help LinkedIn better cater to students' needs and improve the student experience on the platform.

My team and I conducted over 200 survey responses and 50 user interviews to learn about students' professional development needs. We used our findings to create 3 prototypes for new Mentorship, Jobs, and Feed features, and conducted market research to develop a go-to-market strategy for this feature set. Throughout this process, we've also received mentorship and input from 15 generous LinkedIn professionals.

We published a whitepaper, pitch deck, video, solution prototypes, and a Medium case study about the product development process, and pitched our ideas to Tomer Cohen (Chief of Product at LinkedIn) in March 2021. Show less

Co-wrote detailed domestic and international business plans for a company seeking to aggregate autoimmune disease data and patient records from hospitals, pharmaceutical companies, insurance firms, and academic institutions to create machine learning tools that determine potential diagnoses and treatment plans for patients.

For this project, my team was named as a DECA International Finalist for 2 years (2019 International Business Plan and 2018 Independent Business Plan events) and a… Show more

Co-wrote detailed domestic and international business plans for a company seeking to aggregate autoimmune disease data and patient records from hospitals, pharmaceutical companies, insurance firms, and academic institutions to create machine learning tools that determine potential diagnoses and treatment plans for patients.

For this project, my team was named as a DECA International Finalist for 2 years (2019 International Business Plan and 2018 Independent Business Plan events) and a Semifinalist in the Diamond Challenge Business Concept Competition. Show less

This study developed a biomechanical model that reliably simulates the coronary stenting process and the associated strains and stresses in the artery and coronary stent. The model can be used by doctors to adjust various factors of the stent implantation procedure (e.g., stent placement, shape, or material) to minimize arterial damage and thereby reducing the rush of restenosis, a recurrence of narrowing of the blood vessel, due to coronary stenting.

For this project, I was named as a… Show more

This study developed a biomechanical model that reliably simulates the coronary stenting process and the associated strains and stresses in the artery and coronary stent. The model can be used by doctors to adjust various factors of the stent implantation procedure (e.g., stent placement, shape, or material) to minimize arterial damage and thereby reducing the rush of restenosis, a recurrence of narrowing of the blood vessel, due to coronary stenting.

For this project, I was named as a 2019 Regeneron Science Talent Search Top 300 Scholar. Show less

This study developed an accurate poroelastic finite element (PEFE) models to supplement magnetic resonance images to evaluate mechanical tissue damage due to brain tumor growth, which is responsible for many of the symptoms brain tumor patients experience. This project demonstrated that there is a large untapped potential for the combined use of mathematical modeling and MR images to evaluate mechanical brain damage due to tumor growth.

This project was a 2017 Intel International Science… Show more

This study developed an accurate poroelastic finite element (PEFE) models to supplement magnetic resonance images to evaluate mechanical tissue damage due to brain tumor growth, which is responsible for many of the symptoms brain tumor patients experience. This project demonstrated that there is a large untapped potential for the combined use of mathematical modeling and MR images to evaluate mechanical brain damage due to tumor growth.

This project was a 2017 Intel International Science and Engineering Fair (ISEF) Finalist. Show less