Feb 2024 – Jul 2024 · 5 mos

• After a dream ride of nearly 13 years at Netflix, I decided to take an exit from this phenomenal company. I had worked on some very interesting technical problems with some of the most brilliant and thoughtful people that I knew. Given the radical transparency in the company and the projects I worked on, I had the thrill of having front row seats to see how Netflix operates a wildly successful business.
• I left Netflix to challenge myself, to make myself uncomfortable. I figured that the most I can grow is by starting my own company. So, I explored several ideas in domains that I had an interest in. I met with dozens of potential co-founders to learn about the problems that they were considering. Finally in July, a brilliant, thoughtful and conscientious small team of people found me. This engagement checked all my boxes - a great team, a problem that is very interesting to me, a problem that my skills are pertinent to, a problem that creates meaningful societal impact and one that we can build a wildly successful business on.

Mar 2018 – Feb 2024 · 5 yrs 11 mos

• My group was responsible for all the services, algorithms and applications that were required to run and operate the CDN.
• The Control Plane team is described in section below.
• The CDN administration platform supported a very diverse set of functions – hardware development, hardware assembly, inventory control, shipping and logistics, data center planning and rack elevations, partner relationship management, server and network device configuration and monitoring, ISP-facing configuration, monitoring and alerting, software deployment to CDN servers, and more. This fairly lean engineering team of highly talented engineers built a bespoke administration platform that allowed my Operational peer teams to be very lean. The platform administered 20000+ servers and network devices at 1000+ disparate networks in 100+ countries.

Mar 2018 – Feb 2024 · 5 yrs 11 mos

• Part of the leadership team on Netflix's Account Sharing project from early stages to success. I led a cross-functional team that identified, prototyped and implemented networking-based signals that significantly improved household detection accuracy. Engaged with key ISP partners to fine tune networking topology, while minimizing user friction across diverse network environments. Worked with product, legal, public relations and other engineering teams to balance accuracy, usability and privacy. Played an active role in evolving Netflix's pricing and plans, aiming to fairly increase revenue while managing brand risk. Engaged complex solutions with Netflix's C-suite executives, facilitating informed decision-making.
• I worked closely with a small part of my team in the CDN group, to run systems that sourced and supplied context for content catalog identification, plan selection and every other geo-related use case across Netflix. Since Netflix's global expansion in 2016, we actively engaged in the cat-and-mouse game of detecting and preventing VPN use, specifically those VPNs that were directed at circumventing legal geo restrictions. This was critical for maintaining trust with content licensing partners as well as protecting revenue. To address genuine privacy concerns and access issues in restrictive regions and networks, we built out the capability of serving a global content catalog for VPN users. Over the years, as use cases evolved, my team developed several Geo and VPN detection signals and machine learnt models to enhance detection accuracy. Worked closely with public relations, content protection, customer service, product groups and in some projects with C-suite executives.

Aug 2014 – Mar 2018 · 3 yrs 7 mos

• Netflix's CDN has had unparalleled efficiency. A core pillar of this efficiency had been predicting content popularity, pre-positioning content in the most opportune locations and directly steering playback sessions to one of 1000s of locations - all functions that the Control Plane team had been directly responsible for. Our systems constructed and operated on a large part of the Internet's routing table to make steering decisions. We balanced traffic across heterogenous hardware by placing content optimized by linear programming. Files were downloaded onto CDN appliances using a no-master consistent hashed approach. When the pandemic hit and demand for Netflix streaming skyrocketed, we built temporary artificial bandwidth caps so that ISP infrastructure would be able sustain Netflix traffic. This amazingly talented team solved problems of scale, efficiency and resiliency with smart engineering and data science and came up with several radical new inventions saving millions of dollars year-over-year for the company. By delivering content from within the ISP’s network data centers (often >95%) we saved a lot of money for ISPs around the world and helped deliver the highest quality of experience for Netflix users.

May 2011 – Aug 2014 · 3 yrs 3 mos

• One of the founding engineers of Open Connect, Netflix's CDN, responsible for server side development. Designed and implemented the service that CDN appliances interact with. Maintained the routing layer used to direct Netflix devices to particular CDN nodes / partner CDN taking into account content availability, BGP routes, geo proximity and node health. Maintained the content engine that directed respective CDN nodes to fill content based on content popularity and sharding. This involved deciding where to fill from - other CDN nodes, partner CDNs or Amazon S3 based on network topology, content availability, health and proximity. Designed and built a batch-processing compute farm in the cloud that executed tasks via crontab. Researched and identified several advancements in content popularity predictions that is used to lay out content on the appliances.

Apr 2005 – May 2011 · 6 yrs 1 mo

• At Yahoo!, I worked on high scale ad serving, implementing machine learning models for ad targeting and prototyping dynamic ad formats as part of my first stint. In my second stint, I drove the architecture and development for universalizing platform widgets that showed up on every Yahoo! Page. This drastically reduced global update times from months to minutes, marking a significant leap in operational efficiency and unlocked a new revenue stream.