Prabhat Rastogi Email & Phone Number

India, IN

Miami, FL, US

Established the data function in Novo and scaled to 15 analysts, data scientists and data engineers while supporting the business through multiple funding rounds where valuation increased from $120 Mn to $700 Mn.Roles:Head - Machine Learning & Data Engineering (Jan 23 - present)Lead Data Scientist - Customer Retention, Acquisition and Fraud (Sep 21 to Dec.

New York, NY, US

• [Payroll of Xtage Ltd. - A company servicing Oliver Wyman with data hiring on contract] Responsible for end-to-end delivery of modelling &data product development for the client (Oliver Wyman)
• Developed and deployed credit loss prediction CCAR / CCEL models with US bureau data using large spark clusters. The IP for the modelling process and automation sold at $3 Mn to a large US bank.
• Managed data pipelines, model development / deployment and backend integration for a web based "loss forecaster" app targeting mid-sized US banks. Tool was purchased by ~700 banks with revenue ~$1 Mn / year

Gurugram, Haryana, India, IN

• Responsible for all data driven revenue enhancement processes includingtargeted discounting, live overbooking and fraud prevention
• Implemented OYO's first user-segment*hotel discounting model thatfactors in history, real-time pricing, location factors etc. to optimise forROI. Helped increase realised value by $2 Mn / month
• Built and operationalised a gradient boosting based cancellationprediction model pipeline that brings in over $1 Mn / month inoverbooking;
• Optimised the algorithms and processes around partner fraud detectionthat helped bring in over 0.5% in occupancy ($1 Mn+ / month)

San Francisco, California, US

• Managed a team of analysts working on fraud, rider/driver churn, safety/quality incident prevention etc.
• Ideated and created metrics, MIS; quantified impact and mined for insights to build execution plan focused on achieving rider loyalty through "magical experiences" as part of core group led by CEO
• Built text classification models for reducing "time-to-respond" for sensetive safety tickets. Achieved reduction in response times from 6 hrs to <30 minutes on targeted cohort of incidents
• Reduced safety/quality related rider churn by over 20% by building ML models for 'bad' incident preventionthrough differential quality incentivisation, dispatch restriction &targeted driver training
• Developed loss mitigation strategies that helped reduce rider payment fraud, GPS fraud etc. -$2 Mn/year saved

San Francisco, California, US

• Led product and process interventions for driver and rider safety on the Uber platform for India, Bangladesh &SL
• Launched data driven real time dispatch control models to prevent quality/safety incidents (~10% drop)
• Instituted in-person and digital trainings (based on driver clustering using historical data based predictionmodels) on gender sensitization and safety/quality that touch ~70k drivers monthly across 35 hubs pan India

New York, NY, US

• Managed client engagements with a large US bank around enhancing data driven risk management capabilities using advanced analytics and automated decisioning
• Led a team of six analysts; managed end-to-end implementation of fraud mitigation strategies for a top US bank -$10 Mn/year saved in fraud losses
• Built a tool for automated logistic model creation - data preparation (outlier treatment, missing value imputation), feature selection and model validation. Helped reduce time to production for standardised fraud.
• Implemented Random Forest, Gradient Boosting based models on identity theft and managed the implementation across stakeholders from customer experience, risk and tech.
• Received 4+ rating (out of 5) across projects and recognized for effort with client bonus, on-site engagement & promotion in 1 year

New York, NY, US

• Enhanced and benchmarked data driven and model based fraud detection and monitoring system for global credit card portfolio of American Express worth $3 Tn / year
• Implemented financial industry's first machine learning based fraud risk management solution - $15 Mn/year fraud losses saved; This model was pivotal in getting US regulators on-board to approve more complex modelling.
• Explored and validated new machine learning approaches - tree based, support vector, segmented logistic, anomaly detection etc - for card fraud modelling. Based on this, existing fraud detection processes were moved.

Gurgaon, Haryana, IN

- Quantified the social impact & investments required for key state infrastructure projects & supported presentation to UP chief minister with recommendations- Analysed of asset class mix, location parameters for evaluation and optimization of returns from real estate resources.

Brisbane, Queensland, AU

• Estimated the coefficients of ‘Taylor Rule’ for Australian economic data and compared its predictions with observed values
• Modelled federal fund rates and extracted various dependencies using statistical methods
• Received recommendation for adept problem solving skills, initiative and professional approach

Pgp (Mba), Business

Dual Degree (B Tech + M Tech), Microelectronics