Jay Soman Email & Phone Number

California, United States

Fremont, California, United States

• Leading the outlier detection team at Neocortex, managing a Data Scientist to solve complex anomaly detection problems.• Currently working on building an internal solution to detect anomalous reimbursement receipts for a large pharmaceutical client.• Developed and deployed an analytics dashboard for client in Power BI that Internal Audit SVPs are using to track the company expenses

Fremont, California, United States

• Uncovered fraudulent vendor transactions by leveraging graph based outlier detection for pharmaceutical industry clients, significantly enhancing their audit and compliance frameworks.• Boosted employee attrition model’s F1 score by 20% through advanced feature engineering and optimization, elevating prediction accuracy.• Created real-time monitoring dashboards in Tableau focusing on explainable AI, utilizing SHAP for in-depth attrition insights, improving decision-making.• Designed a two-tier purchase order classifier employing BERT using HuggingFace, achieving 92% accuracy to aid the client’s budgeting process.• Leveraged SQL to extract and shape datasets from client databases for key projects like employee attrition, showcasing adeptness in data extraction and preparation.• Developed an internal stats tool with pandas-profiler for quick, comprehensive dataset analysis for new projects, enhancing project efficiency.• Facilitated cross-functional problem-solving with Risk, Product, Marketing teams, showcasing teamwork and adaptability.

Boston, Massachusetts, United States

Data Scientist in the Health Team• Improved customer experience by tuning the ”strain score” algorithm to detect low heart-rate workouts• Increased model sensitivity by 90% for low heart-rate workouts• Enhanced specificity of XGBoost classifier by 3% by removing inaccurate class labels in time-series ECG data• Deployed data pipelines for collection and pre-processing of training data• Created various training/test datasets by writing efficient queries in PostgreSQL• Presented timely analyses of the impact of my projects to senior leadership and stakeholders in the company

Boston, Massachusetts, United States

Project - Sleep Detection Using Wearable Computing• Improved training data quality of sleep detection algorithm by filtering 100s outliers• Increased model sensitivity by 2% by correcting labels of training data

Mumbai, Maharashtra, India

Data Analyst in the Load Forecasting team• Cleaned, processed and set up data pipelines for time-series data• Created dashboards and presentations in Tableau and presented to stakeholders• Contributed in the building of load forecasting models based on the ”similar day” approach• Improved existing prediction models by feature engineering geographical parameters such as ”Discomfort Index”which improved accuracy of one day-ahead load forecasting• Performed model evaluations and contrasted model performances of various models to capture weekdays, weekendsand holiday MegaWatt usage pattern• Worked on capturing new COVID pandemic power usage patterns into additional models

Master Of Science - Ms, Data Analytics Engineering, 3.8

Courses Completed: 1. Machine Learning (PhD level) 2. Deep Learning (PhD level) 3. Data Management for Analytics 4. Deterministic.

Master Of Engineering - Meng, Digital Electroinics

Relavent Courses: 1. Data Coding Techniques for Communication and Storage 2. Advanced Computer Systems 3. System Design