Jul 2020 – Dec 2022 · 2 yrs 5 mos

• I am the software engineering team lead in Alexa Speech Recognition (ASR) Cloud team in Amazon. ASR Cloud team owns Tier-1 services used for speech recognition for all Alexa. Currently leading a team of 10+ engineers for the improvement of speech recognition for Alexa customers to increase adoption of Alexa world-wide. As an engineering team lead I wear multiple hats time to time - from being a developer, a code/design reviewer, a mentor/guide, a SRE to promote operational excellence, on-call for the front line support of a Tier-1 service, an inventor to continuously evolve ASR's architecture, to a planner for current as well as future projects. My key contribution till date are:
• Simplified ASR Cloud services architecture by introducing cell based architecture to develop a low maintenance cost solution to manage fleets for audio processing services containing 10,000+ hosts.
• Introduced ability to conduct large scale A/B testing involving millions of customers for ASR ML model releases which provided much needed visibility on the customer facing metrics.
• Introduced uniform audio delivery latency metrics to triage/debug and gaining insight in delays due to complex audio delivery across entire Alexa stack.
• Designed and developed ML model release automation framework to safely release 100s of ML models per month across multiple audio processing tier-1 services in ASR org.
• Designed and developed false traffic suppression service for Alexa to safe guard Alexa from media induced triggers from millions of devices waking up simultaneously due to commercials.

Nov 2017 – Jun 2020 · 2 yrs 7 mos

• I am the software engineering team lead in Alexa Speech Recognition (ASR) Cloud team in Amazon. ASR Cloud team owns Tier-1 services used for speech recognition for all Alexa. Currently leading a team of 10+ engineers for the improvement of speech recognition for Alexa customers to increase adoption of Alexa world-wide. As an engineering team lead I wear multiple hats time to time - from being a developer, a code/design reviewer, a mentor/guide, a SRE to promote operational excellence, on-call for the front line support of a Tier-1 service, an inventor to continuously evolve ASR's architecture, to a planner for current as well as future projects. My key contribution till date are:
• Simplified ASR Cloud services architecture by introducing cell based architecture to develop a low maintenance cost solution to manage fleets for audio processing services containing 10,000+ hosts.
• Introduced ability to conduct large scale A/B testing involving millions of customers for ASR ML model releases which provided much needed visibility on the customer facing metrics.
• Introduced uniform audio delivery latency metrics to triage/debug and gaining insight in delays due to complex audio delivery across entire Alexa stack.
• Designed and developed ML model release automation framework to safely release 100s of ML models per month across multiple audio processing tier-1 services in ASR org.
• Designed and developed false traffic suppression service for Alexa to safe guard Alexa from media induced triggers from millions of devices waking up simultaneously due to commercials.

Jun 2015 – Oct 2017 · 2 yrs 4 mos

• I was part of Catalog Metrics team in Amazon. Catalog Metrics team is part of ASCS (Amazon Selection and Catalog Systems) organization. My projects include, a pipeline for publishing catalog metrics, a platform to generate catalog metrics and a user’s dashboard. Internal teams at Amazon use metrics publishing pipeline to publish metrics generated by them. User’s dashboard is used by leaders and category managers in Amazon world-wide. Category managers and leaders use this dashboard to visualize metrics related to their part of the catalog. It helps them in taking business-critical decision for improving and expanding Amazon’s catalog. Platform for generating metrics can be used to generate important metrics from catalog activity feeds of various teams at Amazon. I was lead engineer on the projects and responsible for all deliverable from architecture, development and deployment.

Jun 2013 – Jun 2015 · 2 yrs · Hyderabad, Telangana, India

• Compliance development team works toward achieving automation for most of the possible piece of task for the Firm's Compliance team. It constitutes not just one of the major decisions making part in the day to day trading activity of the Firm, but also ensuring compliance with various regulatory laws imposed by regulators globally. The development team is responsible for providing Compliance users a real-time platform for monitoring, surveillance and historical analysis. My projects majorly include Compliance Platform, insider trading surveillance, anti-money laundering check, best execution, restriction reporting for IAM trading, Position Limits monitoring and swap reporting – dodd-frank act.

May 2011 – Jun 2013 · 2 yrs 1 mo · Hyderabad, Telangana, India

• Compliance development team works toward achieving automation for most of the possible piece of task for the Firm's Compliance team. It constitutes not just one of the major decisions making part in the day to day trading activity of the Firm, but also ensuring compliance with various regulatory laws imposed by regulators globally. The development team is responsible for providing Compliance users a real-time platform for monitoring, surveillance and historical analysis. My projects majorly include Compliance Platform, insider trading surveillance, anti-money laundering check, best execution, restriction reporting for IAM trading, Position Limits monitoring and swap reporting – dodd-frank act.

May 2010 – Jul 2010 · 2 mos · Hyderabad, Telangana, India

• I worked on project to reduce false positives reported by Anti-Money Laundering check tool and performance improvement of a periodic exhaustive anti-money laundering check. I searched and evaluated new phonetic algorithms to reduce number of false positives and performance improvement. As a part of my task, I utilized data indexing techniques to improve performance of existing system 2 to 4 times.