Sep 2024 – Dec 2024 · 3 mos · New York, United States · Remote

• >> Designed and implemented an end-to-end pipeline using multimodal AI to extract insights from complex PDFs, combining visual (charts, tables) and textual data for automated analysis.
• >> Iteratively evaluated models including CLIP, BLIP, ViT, and successfully deployed a MiniCPM-V multimodal model for integrated image-text reasoning and structured chart interpretation.
• >> Delivered a scalable solution that reduced manual effort, enhanced insight accuracy, and gained buy-in from the Data Science team for production deployment.

Jun 2024 – Aug 2024 · 2 mos · New York, United States · Hybrid

• >> Worked in 2nd Line of Defense (2LOD) to build a challenger model for regulatory compliance; deployed an Anti-Money Laundering model using XGBoost to enhance Sanctions Screening, automating the closure of low-risk alerts & prioritizing high-risk ones, increasing screening efficiency by 30%
• >> Applied advanced feature engineering techniques to address imbalanced class problems, leveraging insights from Fraud Detection models. Developed a feature selection framework using XGBoost and SHAP importance, enhancing model interpretability and scalability

Jan 2024 – Dec 2024 · 11 mos · New York City Metropolitan Area · On-site

• >> Oversaw and administered accommodated exams for students registered with Columbia Health, Office of Disability Services, ensuring full compliance with ADA guidelines and academic policies.
• >> Maintained a secure, distraction-free testing environment while safeguarding academic integrity and handling sensitive materials with discretion.
• >> Coordinated exam logistics—including extended time, assistive technologies, and alternate formats—and collaborated with faculty and support staff to meet individualized student needs.

Jun 2022 – Jul 2023 · 1 yr 1 mo · Hybrid

• >> Brazil Demand Forecast Optimization: During my onsite visit to the São Paulo/Brazil Office of AMBEV, I identified critical gaps in the Disaggregation logic that affected the market - particularly during PINC weeks. Later collated complete datasets for 2017-22 and established the initial logic for testing in consultation with Demand Planners. Also incorporated the use of the Revenue dataset to extract PINC information and initiated communication for future projections from the Sales Teams.
• >> Soccer Event Encoding: Introduced Soccer Event flags into o9 statistical models, which significantly enhanced forecast quality, especially during events like FIFA. This led to more accurate market forecasts during match weeks, aiding in better inventory planning.
• >> Hybrid Model Ensemble Selector: Played a pivotal role in implementing a new model selection methodology for o9 models in Brazil, addressing implementation issues and pipeline delays. I analyzed the existing forecasting process and proposed improvements based on business knowledge. This technique was later also adopted for models in Mexico and Colombia.
• >> Improving Feature Data: Designed auto-refresh APIs for macroeconomic and weather data in o9 models, working closely with Demand Planners and third-party data vendors. This resulted in improved forecast accuracy upon testing with Soccer, Price, Holiday, and Google Mobility datasets.
• >> South Africa Forecasting Improvement: As the leader of the o9 Model Development team in SA, successfully reduced over-forecast bias by approximately 10% and improved forecast accuracy by around 2.8% within a strict 4-week deadline.

Feb 2021 – Jun 2022 · 1 yr 4 mos · Hybrid

• >> MAZ Data Pipeline Enhancement: Guided the MAZ and o9 teams in resolving preprocessing issues and proposed improvements based on business knowledge. These enhancements later became part of the Continuous Improvement Initiative for MAZ models.
• >> Empties Returns Forecast: Provided consultancy to the Logistics Analytics team, helping them develop a scalable o9-driven solution using ML to forecast empties return volumes based on historical sales and seasonality patterns.
• >> South Africa Continuous Improvement: Collaborating with various teams, I worked on recalibrating the existing model repertoire to improve Forecast accuracy. This included adding capabilities for OOS treatment, Homologation, and Outlier correction.
• >> EUR Demand Planning (UK Pilot): Played a key role in establishing the Demand Planning Framework for the UK, which is was later replicated by o9 for the entire EUR zone. Also provided technical guidance to EUR's Demand-Driven Supply Program.

Jul 2019 – Jan 2021 · 1 yr 6 mos · Bengaluru, Karnataka, India · On-site

• >> Contributed to the development and validation of critical IFRS9 models, including PD, LGD, and EAD, providing key estimates for VaR, PE, and RWA management, while also designing statistical simulation tools for precise risk measurement and automating monthly Data Risk reports for multiple card segments.

May 2018 – Jul 2018 · 2 mos · Bengaluru, Karnataka, India · On-site

• >> Conducted comprehensive analysis of Small Business Lending portfolio, evaluating performance metrics and risk mitigation strategies over time.