Feb 2025 – Jul 2025 · 5 mos

• Make AI chatbots personalized so they are helpful and engaging.
• Improved experimentation velocity at Meta AI by developing infrastructure for training data collection from human, synthetic data generation, prompt centralization and evals.
• Improved engagement by fine tuning llama through SFT, DPO, RLHF and reward model training

Feb 2023 – Feb 2025 · 2 yrs

• Deliver on Facebook's obligations towards data use and privacy regulations while minimizing privacy tax for the developers.

Sep 2018 – Jan 2023 · 4 yrs 4 mos

• I develop systems, consult across product engineering teams, ensure engineering design and operational excellence, identify opportunities of impact and build a world class product. The systems protect Facebook products from various kinds of bad content and actors.

Jul 2017 – Sep 2018 · 1 yr 2 mos

• I have been developing algorithms that detect audio/video copyright infringement on facebook and instagram. I work with the product teams to define the problems and use machine learning techniques for them. It includes literature review, training and collaborating with teams around the world for data collection, developing accurate and highly performance optimized models(performance is important at facebook scale), integrating them with facebook infrastructure and repeating.

Jan 2017 – Jun 2017 · 5 mos

• I worked on the videos infra team. I worked on Livewith which allows two people to be live together on facebook surfaces. Specifically, I worked on compositing the two media streams, which involved working on streaming protocols and tools like ffmpeg and gstreamer.

Jun 2014 – Dec 2016 · 2 yrs 6 mos

• As a backend engineer on Real Time Infrastructure, I built a set of distributed systems that enable all devices/browsers running Facebook’s leading products to stay connected to our backend 24x7 and be engaged in real-time exchange of content, push notifications and signaling. The services handle close to a trillion messages a day, solving one of the biggest scale problems in the world.
• In the earlier part, I focussed primarily on the gateway layer which is essentially the last mile to the device and implements the protocols for communication like MQTT, HTTP and XMPP.
• Next I worked on the presence system which is responsible for the *green dot* or availability status on all facebook surfaces. It's an in memory system that handles over a billion RPC requests a minute, does complex privacy checks, staying up to date with a large in-flow of ephemeral user availability events.

Jul 2012 – May 2014 · 1 yr 10 mos

• I worked as a backend systems engineer on Ads and Pages insights platform. The insights platform is a distributed system that collects data from various sources across the site, aggregates it for easy consumption and presents it to the businesses, all in real time. The data provides insight into the perceived quality of their content and measure user sentiment/engagement across various demographics and drives key business decisions. The platform crunches billions of events per day for millions of active pages and advertisers.

Sep 2024 – Jul 2025 · 10 mos · Greater Seattle Area

• I hold a portfolio of 5 startup companies.
• Independently:
• Cielara AI
• Byteroll
• Incantio
• In a special purpose vehicle:
• Aveera Health
• Image Rehab

May 2011 – Jul 2011 · 2 mos · Bengaluru Area, India