Promoted for my prior work. Applied LLMs at Pinterest.Part of the Engineering Leadership Team.Converged teams to a stable state with happy engaged leads. Left Pinterest with good vibes - It's time for me to do something new.

Led the delivery of Pinterest’s stream processing platform on Apache Flink:Pinterest's streaming infrastructure is the winner of the Confluent 2022 Data Streaming Awards for the best company wide data streaming implementation. - 200+ engineers using the platform internally.- Grew from ~$1M to become the 7th largest platform at Pinterest and operating at $xxM in resource spend and a 20M+ msgs/s scale. - Achieved 3x faster time to market for comparable jobs vs industry baselines, lowered spends per job by 50% and version upgraded the entire platform through 3 cycles in 3 years. - Exponential adoption: 0 to 5 clients in 6 months; 60+ clients in 24 months. 100+ clients in 36 months. Business critical impact generated across all Pinterest engineering orgs.As a senior TL, I either led or contributed to:- Build vs Buy decisions, Org level annual strategic roadmaps, Cross Org Cost Saving collaborations, Architecture review and design, additional Infra Spending justifications and Team mentorship & productivity.- I'm directly responsible for a single digit % increase in Pinterest's profitability. - A 15 person team makes all of the above possible.

Bootstrapped the ML Platform team & stabilized the Signal Platform team at Pinterest. - Set the ML Platform Strategy in collaboration with clients, partners and EMs. - Built out the Realtime ML Monitoring infrastructure that continues to monitor and protect all major models & features on the main Pinterest surfaces (Homefeed, Ads Ranking, Ads Retrieval, Search, Content Quality).- Led the Signal Platform team through a challenging period of high operational load and a critical migration to a new platform. Instituted systems, processes and productive approaches. The team achieved a +40 increase on their NPS score with clients.- Founded the Near Realtime Galaxy project that ultimately became the seed for Pinterest's Stream Processing Platform and powered their Pinterest 2.0 roadmap.

LinkedIn Feed Infrastructure:Designed and Built a near-realtime Comment Relevance system that powers all relevant comments on LinkedIn based on a machine learnt model and a near-realtime indexing and serving architecture. +22% Comment views iOS, +14% Comment views Android, +14% Social Gesture Uniques on Comment Replies, +14% Viral actions from Comments. (See links for architecture and press coverage)Built a non-parametric MDMC model for distinguishing creators and consumers on the LinkedIn feed. (patent granted).Flagship Global Initiatives: Worked cross org. Led teams of 2-4 engineers.Helped pivot the LinkedIn feed experience around Groups, HashTags and Recommended content.Restructured FollowFeed's internal query language to support parallel, overlapping A/B experiments across products. The restructured query language supports all LinkedIn feed traffic across 130+ use cases on the website and comes with an invariant verification test-suite (2000+ tests).

2013H2 - 2016 (Search Infrastructure - Recommendations):Collaborative filtering. Built systems that process > 1 billion requests a day in the critical path of Google's Search and Mobile experiences.- Twitter recommendations for queries on Search: Powers all non-author Twitter results on the search results page in the critical request path (low latency, high availability, globally distributed indexing + serving). Press: https://blog.twitter.com/2015/a-new-way-to-discover-tweets http://googleblog.blogspot.ru/2015/05/tweets-take-flight-in-google-app.html- Shared offline infrastructure for location data (~ petabytes of data). Signal to location based predictions on several Google systems (Search, Android: Google Now).- Eval and review infrastructure for recommendation relevance (used internally by 30+ teams).- Several internal infrastructure projects.2011 - 2013H1 Social team (Google+):- Unify Google look and feel across properties with a common infrastructure. 2 UI refreshes.- Backend endpoints for product access customization on all Google properties (100+ products)- UI and infrastructure features around authentication (used by 1 billion+ users).- 10+ new products onboarded.

As the Chief Technical Architect of Zigndog, I was responsible for managing and defining client requirements, production deadlines, code quality and the complete architecture of the solution delivered by the company. I was also responsible for mentoring other team members, providing timely and highly technical in-depth advice and generating new clients. I helped set up the complete tech stack for the company from scratch.

I'm one of the 4 interns from India selected to work on projects at Qualcomm HQ at San Diego as part of the Qualcomm Technology and Innovators Program. I've worked on I2C interfacing of a low power PAN/BAN chip as part of the Corporate R&D group.

BizTalk Server Developer Tools Team. I implemented a novel, stateless "coalescing" algorithm on the .NET TreeView Control to implement a feature proposed by Robertson et al. in their HCI Paper - Visualization of mappings between Schemas - Conference on Human Factors in Computing Systems, ACM SIGCHI.

Responsible for the complete organization of an International Coding Event - Insomnia featuring over 70,000 page hits. Handled sponsorship commitments, technical requirements and publicity and promotion of the entire event. Was closely involved with the sponsors of the fest. Also managed the Cognizance website, which became India's most visited technical fest website as per Alexa's rankings during Feb-March 2009 with nearly 1.5 million page views for 2009.

Led a team of 300 dedicated students in organizing over 50 events benefiting the underprivileged during a period of about 8 months. Streamlined communication within the organization and promoted documentation of activities via photos and articles. Introduced new events and educational priorities. I was a student advisor to the team during 2009-2011.

Development of a online collaboration application for one of their most important clients using AJAX and ASP. Automation of office tasks via VB .NET and VBA. Development of a web robot for fetching and classifying the entrance examination results of students - in Python and PERL.