During my tenure here, I have led the technical direction on these teams: Fulfillment, Reliability, Logistics and Platform. Most impactful projects are:Architect of Postmates matching/routing algorithm. It is a highly scalable and extensible system supporting advanced routing like batching (multiple order, same merchant) and chaining (multiple order, nearby merchants). It is the industry leader in courier efficiency (nearly 3 trips / hour in strategic markets). Fun fact: It generates over 64 different types of dispatches today.Architect of Postmates dynamic pricing system. Responsible for maintaining a balance of supply and demand in the marketplace by blitz pricing, fleet throttling and heat maps.Other projects that I have led:• Scaling matching algorithm to support 10x ongoing deliveries.• Migration of Fleet/Logistics services from AWS to GCP.• Proactive Live Order Support System that detects if a courier is off-track during an ongoing delivery and takes appropriate actions.• Fleet Reliability Service which is an event driven system responsible for real-time fleet throttling and suspension.

Developed audio and voice software for Multimedia Team• Optimized and implemented fwk for features like Voice activation ("Ok Google!"), Noise Cancellation, etc.• Optimized performance of Speaker protection algorithm by 22%• Re-designed resource voting (clks, bw) giving finer control to individual streams using a per-port scheme.• Designed and implemented framework for resampling fractional sampling rates on Audio.• Re-designed framework to enable 3rd party module support.• Main POC for commercialization of MSM8994 and MSM8992 chipsets for multimedia audio.

Fast facial expression recognition based on local binary patterns• Developed a system that recognizes human expression like Happy, Sad, Surprise, Anger, etc at 4.8 FPS.• Achieved 86.7% detection efficiency using Adaptive boosting of extracted features.• Reduced time to process an image by 84% using binary bitwidth reduction and Support Vector Machine.Binary pattern based edge detection for motion estimation in H.264/AVC• Proposed a new low complexity binary approach for block based motion estimation.• Treats pixels in a group and detects edges that contributes most to distortion function.• Reduced complexity by 62% as it only uses compare operator while others use arithmetic operators.Breast Cancer Recognition and English Characters Recognition using Neural Networks• Improved breast cancer detection rate from 93.86% to 96.73% by using modular neural network.• Implemented a real-time clustering algorithm using Adaptive Resonance Theory (ART).• Extended ART clustering algorithm to automatically map and detect characters degraded with random noise.

Worked in the Windows Team (WEST) responsible for developing and characterizing performance of Mobile Station Modem• Remodeled BAT and Stability MS Test Central suites on Snapdragon 8x30 and 8x60 chipsets.• Handling automation of seamless onboarding of multiple new devices on the rack.

Developed a Real-Time Log Management tool (QRadar) to efficiently analyze network and security event logs• Performs ad-hoc log parsing and generates an actionable correlated aggregate schema for it.• Indexing suspicious fields enables us to quickly detect, manage, or even predict security threats.

Developed an absolute encoder which provides a contactless method for displacement measurement• Achieved 98% accuracy in the detection of movement with accuracy reaching smaller than 1 micron.• Used for Displacement Measurement and Positioning of autonomous robots with Optical Sensors.