Sanjana Srivastava

Sanjana Srivastava

ML Engineer at Trase ai

I am a Machine Learning Engineer at Trase ai, specializing in deep learning, computer vision, multimodal learning, and responsible AI product development. Academically, I hold a Master’s in Data Science from the University of Virginia, and an integrated Master and Bachelor of Technology in Geological Technology and Mathematics from the Indian Institute of Technology Roorkee.

Previously, I collaborated with Deloitte to build LLM evaluation pipelines and served as a Research Scientist at Univesity of Virginia with Dr. Donald E. Brown and Dr. Sana Syed, where I trained ML models for disease classification and pattern recognition in multimodal medical data. Previously I have interned at BNY Mellon and ZestMoney as a Software Engineer.

📖 Education

  • 2024 - 2025: Masters in Data Science | University of Virginia
  • 2021 - 2022: Masters in Technology | Indian Institute of Technology, Roorkee
  • 2017 - 2021: Bachelors in Technology | Indian Institute of Technology, Roorkee

💻 Experience

  • Machine Learning Engineer at Trase ai | August 2025 - Ongoing
    • Architecting and deploying machine learning solutions to automate complex data workflows and processing pipelines across the oil & gas and healthcare sectors.
  • Machine Learning Researcher at University of Virginia | July 2022 - May 2024
    • Worked with a multi-disciplinary team of medical professionals and engineers to study gut functions.
    • Implemented novel Deep Learning models for disease diagnosis and quantification.
    • Leveraged Machine Learning for pattern recognition in tissue images, clinical data, and transcriptomic data.
  • Software Engineering Intern at BNY Mellon | Summer 2021
    • Implemented functional and unit testing for internal applications.
    • Developed a custom XML to CSV Parser Utility.
    • Conducted A/B testing to assess performance optimizations in internal applications.
  • Software Engineering Intern at ZestMoney | Summer 2019
    • Implemented a Payment Gateway at checkout using SpringBoot & MySQL database.
    • Developed a custom user Signup interface using Retrofit (Android) with MVVM architecture.
    • Improved code usability in a high-paced fintech industry startup environment.

📂 Projects

Evaluating efficacy of synthetic images generated using diffusion models
Developed diffusion models to generate histology patches conditioned on nuclei locations. Validated the use of synthetic patches for improving downstream segmentation and classification tasks.
Code

Deep learning based detection and visual understanding of diseases using medical imaging data
Performed patch-based invasive ductal carcinoma (breast cancer) and gastrointestinal disease detection. Implemented convolutional neural networks (CNN) for classifying whole slide images and biomarker data. Implemented Gaussian clustering methods to identify recurring visual patterns in diseased biopsies.
Code

StyleSwap: Text-driven latent diffusion model for localized fashion image editing
Designed a localized fashion image editing pipeline using text-driven latent diffusion and semantic segmentation.
Code

Humorous Image Captioning System
Implemented a self-attentive encoder-decoder framework to generate humorous captions for images indistinguishable from human generated memes.
Code

Correlating disease gene signature with imaging data
Designed a deep learning framework to identify image features associated with functional gene clusters. Identified important gene signatures and their correlation with visual patterns in biopsies.
Code

Deep learning based semantic segmentation on brain MR images
Performed tumor segmentation using a U-Net architecture on MRIs. Used PyTorch for training model on MRIs from The Cancer Genome Atlas (TCGA) lower-grade glioma collection.
Code

Petrographic characterisation of a chondrite sample
Investigated the mineralogy and major element geochemistry of mineral phases present in the chondrite section. Performed Electron Probe Micro Analysis to obtain backscattered electron images of the sample.

Alzheimer’s disease analyses using patient data
Explored factors associated with Alzheimer’s, developed a predictive model, and conducted statistical analyses using regression models on patient chart data.
Code

Chicken Litte Run
A fun run and dodge game developed on HTML Canvas where the Chicken Little must run and dodge the falling sky.
Code

📝 Publications

Comparative Study of Large Language Model Evaluation Frameworks with a Focus on NLP vs LLM-As-A-Judge Metrics
S. Srivastava, A. Alabdulwahab, C. Japic, C. Le, D. Dubey, D. Trivedi, J. Hope, P. Stone, A. Tashman, A. Zhang.
Systems and Information Engineering Design Symposium (SIEDS) 2025.
Paper

Machine-learning-based integrative–‘omics analyses reveal immunologic and metabolic dysregulation in environmental enteric dysfunction
F. Zulqarnain, X. Zhao, K. Setchell, Y. Sharna, P. Fernandes, S. Srivastava, A. Shrivastava, L.Ehsan, V. Jain, S. Raghavan, C. Moskaluk, Y. Haberman, L.A. Denson, K. Mehta, N.T. Iqbal, N. Rahman, K. Sadiq, Z. Ahmad, R. Idress, J. Iqbal, S. Ahmed, A. Hotwani, F. Umrani, B. Amadi, P. Kelly, D.E. Brown, S.R. Moore, S.A. Ali, S. Syed.
iScience 2024.
Paper

Quantitative Morphometry and Machine Learning Model to Explore Duodenal and Rectal Mucosal Tissue of Children with Environmental Enteric Dysfunction
M. Khan, Z. Jamil, L. Ehsan, F. Zulqarnain, S. Srivastava, S. Siddiqui, P. Fernandes, M. Raghib, S. Sengupta, Z. Mujahid, Z. Ahmed, R. Idrees, S. Ahmed, F. Umrani, N. Iqbal, C. Moskaluk, S. Raghavan, L. Cheng, S. Moore, S.A. Ali, J. Iqbal, S. Syed.
The American Journal of Tropical Medicine and Hygiene (ASTMH) 2023.
Paper