
Engineering leader with 13 years at Amazon, including 7+ years of management experience delivering scalable, customer-focused solutions in streaming ad tech, fashion tech, and enterprise systems. Proven track record of working backwards from KPIs and success metrics by building and leading high-performing teams, launching innovative advertising products, and scaling systems built on AWS.
Lead the Ad Innovation and Inventory Share Monetization engineering charters for Prime Video Ads, driving scale and revenue growth with advertisers as the customers:
Drove ~$XXXM in incremental annual ad revenue for Amazon publishers (Freevee, Prime Video) by leading the development and launch of new monetization products. Delivered high-visibility ad formats—including Interactive Pause Ads, Brand Storytelling segments, Sponsored Slots, and First Impression Takeover —through rapid collaboration with product, content, and marketing teams to meet upfront commitments and accelerate advertiser adoption.
Increased product reliability and reduced operational load by building automated tech systems that streamlined and operationalized complex ad-product setup workflows, using GenAI-assisted inputs and validation, that lowered setup friction and reduced defects, enabling product teams to launch new campaign faster and with higher accuracy especially in Q4 peaks.
Increased linear ad supply and unlocked new revenue opportunities by leading the FAST Linear channel experience across Prime Video FAST, Fire TV Channels, and Prime Video Channels, delivering approximately 10 inventory-share integrations annually. Enabled this scale by architecting advanced inventory controls (category exclusions, competitive separation), and partnering with third-party tech teams to build OpenRTB-compliant ad-server capabilities while maintaining strict viewer privacy protections.
Drove growth and customer engagement for Amazon Fashion’s Personal Shopper subscription by leading a cross-functional engineering team (10 SDEs, 5 QA) responsible for the end-to-end technical architecture, UX flows, and personalization systems that power the subscription experience.
Increased new subscription conversions and reduced onboarding friction by engineering a high-throughput feedback pipeline that captured customer style preferences, enabling photo-upload workflows, and building a promotion-configuration system used during Prime Day and seasonal campaigns. These capabilities improved input quality, reduced drop-offs, and enabled targeted incentives.
Accelerated experimentation velocity and improved insight quality by designing and delivering a scalable UI survey framework supporting dynamic layouts, real-time instrumentation, and A/B testing hooks. The platform was adopted by 5+ teams across Amazon Fashion, enabling richer behavioral data collection and rapid iteration on recommendation and fit models.
Improved data availability and platform reliability for Amazon Business by leading the Data Aggregation Systems team responsible for near–real-time event processing and multi-tier storage for downstream customer and analytical workloads.
Enhanced search performance and scalability by architecting and launching an AWS Elasticsearch–backed search layer, enabling faster retrieval of aggregated business data for customer-facing features.
Expanded Amazon Business’s external integration capabilities by leading the design and launch of its first external-facing API delivering post-purchase ordering and transaction data. The successful beta, adopted by enterprise customers for system-to-system integrations, led to a dedicated product charter, and a full engineering team spin-out.
Increased durability and cost efficiency by designing a multi-tier storage architecture using DynamoDB and Elasticsearch for hot data, with Athena and Redshift Spectrum for warm analytical access—supporting diverse latency, availability, and cost profiles across use cases.
US011869240B1 · Issued Jan 9, 2024 - Semantic Video Segmentation
This patent relates to a machine learning technique for automatically segmenting videos based on different categories (scenes including a particular actor, fight scenes, scenes suitable for children, etc.). Timecodes pertaining to the segments are stored in a segment database and are tagged with category data. Thereafter, a user can request scenes of a particular category and have only those scenes classified as pertaining to the specified category played back without playing intervening portions of the video. Additionally, multiple categories can be specified and only those segments that include all the specified categories are returned. Users are able to provide feedback on the received segments in order to improve model performance.