Nov 2024 – Present · 1 yr 6 mos

• I lead a team building an Intelligence Engine that fuses multi-modal creative assets with disparate marketing data (campaigns, traffic, attribution/conversions etc). This powers our agents to deliver predictive insights, bridging the gap and slashing the cycle time between marketing and creative agencies.
• We are AI native team that design, develop, deploy and maintain ML models and agentic workflows that extract deep insights from creative assets, forecast trends, predict campaign outcomes, recommend high-impact variations to drive autonomous campaign optimization.

Jan 2021 – Nov 2024 · 3 yrs 10 mos

• Built and led a team of 10 Applied Research Engineers who have been responsible for AI/ML initiatives in Adobe's Advertising Cloud. Together, this team designed, developed and deployed features that drive $50M+ ARR
• Working on tools to plan and manage campaigns on Search Ads(Google, Bing etc), Social Ads(Meta, LinkedIn etc) & DSP.
• Building low cost ML Ops pipeline with simple yet efficient development and deployment framework; model and data drift monitoring and alerting.
• Enhancing usability and explain-ability of models with the ability to experiment, integrate and analyze across different functions such as sales, customer support and core engineering. This helped in creating robust and personalized sales pitches, increased adoption of new ML driven features, efficient data/model pipelines and enhanced clarity in ML roadmap.
• Partnering between Product Management, Sales, Core Engineering Teams. Undertaking next gen initiatives such as smart insights, co-pilot for campaign performance, Gen-AI powered creatives
• Some key projects we are working on:
• o Using RL to solve cold-start of an ad campaign.
• o Using Transfer Learning to predict performance of an ad
• campaign
• o Forecasting Conversions using Bayesian Belief networks and
• Gaussian Mixture Models
• o Redesigning MLOps pipeline by modernizing the workflow
• engine into cloud native services and adding frameworks like Blue/Green deployment, A/B test environment etc

Oct 2014 – Dec 2020 · 6 yrs 2 mos

• Researched, Designed, Developed, Deployed and Maintained models that predict Ad-
• Campaign performance, optimally allocate budget across various campaigns.
• Campaign Budget Allocator - Allocate budgets across portfolio of campaigns. Used Constrained Optimization with explore-exploit strategy. Built a pipeline to handle different types of campaigns -text, shopping, smart bidding etc
• Performance forecaster and simulator - Built a tool to forecast the performance of ad campaigns and simulate the effect of budget changes. Designed solutions using algorithms like ARIMA, STL(Prophet), Hierarchical regression, Bayesian Belief Networks, Constrained Optimization, Reinforcement Learning, Genetic Algorithms among others.
• Long term forecasting and budget pacing - Used Time Series Forecasting(STL), Linear Programming and RL to optimally pace budgets across the duration of the campaign. Designed pipeline that can work with Search, Social and DSP campaigns and has the ability to handle special events like promotions.
• Bidding for DSP placements: Built a statistical model to identify any over or underspend trends and adjust bidding strategies accordingly.
• Creative rotation for DSP ads: Built a recommendation engine that identifies the creative using various signal from auction and surfer. Working on building a GenAI pipeline to auto generate best creatives.
• Developed and automated machine learning pipeline that enabled faster model training and deployment. This Canary like deployment framework enabled A/B testing multiple approaches and reduced the time from research to deployment by 50%
• Used word embeddings to build clusters of similar keywords and then forecast the ad traffic on these clusters.