Mansi Pandya

Experience

• 

  The RealReal

  Jun 2022 - Aug 2022

  Machine Learning Engineer

  • Developed a predictive model to forecast the likelihood of member consignment inquiries within the next 30 days based on their activity in the previous 30 days.• Extracted large datasets from GCP using SQL queries and created visualizations using Tableau.• Deployed LightGBM Classifier as the primary model for prediction, while also experimenting with various other models such as Random Forest, Catboost etc. leading to a 20% increase in the F1-score.• Conducted SHAP analysis to interpret model predictions, providing insights into feature importance and their impact on the prediction outcome.• Collaborated with engineering and MLOps teams to establish a pipeline, successfully deploying it to staging on GCP.• Presented insights to stakeholders and cross-functional data science team for comprehensive understanding. Show less

• 

  UC San Diego

  Sept 2022 - Jun 2023

  Graduate Teaching Assistant
• 

  Knidian

  Sept 2023 - Sept 2024

  Machine Learning Engineer and Researcher

  • Engineered a program to extract relevant symptoms from medical history text.• Executed end-to-end development tasks including data cleaning, model architecture design, training, prediction, and evaluation.• Explored five distinct models: traditional word matching, word matching enhanced using embedding, RAG, RAG enhanced using Prompt Engineering and MedCAT for symptom retrieval.• Created vocabulary and concept database models for MedCAT and utilized them to extract symptoms efficiently.• Orchestrated deployment pipelines utilizing Docker and Kubernetes on GCP, and integrated CI/CD tools for seamless deployment processes.• Collaborated with medical professionals to curate a diverse test set mirroring authentic clinical scenarios. Additionally, sought guidance from a UCSD professor for research insights, augmenting the project’s scientific robustness and significance. Show less

• 

  MoEngage

  Nov 2024 - now

  Data Scientist