Jan 2023 – Jul 2025 · 2 yrs 6 mos · San Francisco, California, United States

• Chat platforms today are inherently broken. Slack and Discord, the most commonly used chat platforms, while useful for ephemeral communication, create knowledge black holes, hindering the effective storage and retrieval of information. I have thought about this problem for the past 5 years - Struct is a culmination of those ideas.
• Built a real-time chat platform for Open Source communities, designed around threads, feeds and AI.
• Struct's launch post climbed to #1 on Hacker News for over 6 hours, and was on front page for a day [1].
• Struct is pioneering custom real-time feeds, allowing users to create feeds over channels, tags and users -- while still ensuring real-time updates and unread counters across all these feeds.
• Struct generates a title and summary for every thread on the platform, maintaining context for easy consumption.
• Struct bot uses retrieval augmented generation to answer users' questions based on previous conversations, while enforcing access control. It also allows multi-user / GPT interaction for enhanced collaboration.
• Introduced public to internet channels. Threads in these channel become available as static pages indexable by search engines and GPTs, building a knowledge base from conversations.
• Pioneering dynamic access control where a thread's access could be changed on-the-fly to ensure the right audience.
• [1]: https://news.ycombinator.com/item?id=39557188

Feb 2022 – Jan 2023 · 11 mos · San Francisco, California, United States · On-site

• Built a 10x faster blockchain indexing system, serving Ethereum JSON-RPC APIs and streaming blockchain data. Benchmarked against top players like Alchemy, Infura; and against the official Ethereum node, geth.
• 0xFast system has a 5-10x smaller disk footprint compared to geth or Erigon, popular blockchain node systems.
• Built a new novel system which separates storage from index, designed to take advantage of blockchain data characteristics like immutability and externally achieved consistency.

Oct 2015 – Jan 2022 · 6 yrs 3 mos · San Francisco Bay Area · On-site

• Inspired by the Knowledge Graph system that I had built at Google, I started Dgraph -- an open source, Spanner-like, distributed, ACID transactional, synchronously replicated graph database.
• As a solo-founder, raised 15M in Venture Capital, grew the company to 1M ARR, grew the team to 50+ strong, with 40+ engineers. I ran the entire org as the CEO and acting CTO.
• Dgraph Labs was the youngest company to be featured by Forrester in the graph DB space as an emerging contendor. With close to 10K open source active instances per week, Dgraph had 250+ paying customers, including fortune 100 orgs.
• Fueled by many of my articles which hit the front page of Hacker News, Dgraph became the most "starred" graph-database on GitHub, surpassing all existing players including Neo4j -- over 20K GitHub stars.
• Dgraph was and still is, the only native graph database to have been Jepsen tested for distributed transactional guarantees across various disk, process, clock and network faults.
• As side projects to support Dgraph, we built Badger, an embeddable key-value database in Go as a replacment for RocksDB. Most notably used by NYTimes and IPFS, Badger has 13.5K GitHub stars.
• Also built Ristretto, a fast LRU-cache, with the highest hit ratios compared to all other embeddable caches in Go; and sroar - serialized roaring bitmaps, which outperformed the official Go library by 7x.

Nov 2013 – Sep 2015 · 1 yr 10 mos · Sydney, New South Wales, Australia · On-site

• Moved to Australia early 2014, and started a company called Snapknock (list.cc) -- a social take on craigslist, aimed at generating trust among buyers and sellers. Side note: Facebook launched Marketplace in 2016 aimed at solving the same problems.

May 2013 – Oct 2013 · 5 mos · San Francisco Bay Area · On-site

• Had a short stint at Quora in 2013, where I most notably moved an expensive batch job to run on DB slave entirely, eliminating latency spikes for Quora's end users.

Aug 2007 – May 2013 · 5 yrs 9 mos · San Francisco Bay Area · On-site

• [2011-2013] Started and lead a team of engineers to build Google's first graph indexing and serving system to consolidate and serve all knowledge previously stuck in various one-box backends. The system continued tobe built and iterated upon even after I left, and is still in production today.
• [2010-2011] Built a knowledge engine which could understand user's query and generate answers, which became the basis of Knowlege Bar at google.com. For example, [tom hanks movies].
• [2007-2010] Received an OC award for successful launch of Caffeine: Google's incremental indexing system providing 50% fresher results, and the largest collection of web content ever offered by Google, holding over 100 PB of data. As part of index selection group, I worked on various index launches tackling size control, index quality and page freshness issues.

Jun 2006 – Jan 2007 · 7 mos · San Francisco Bay Area · On-site

• - Interned at Google with the i18n team. My love for systems became obvious when I found an opportunity to work with the Bigtable team, porting Bigtable's C++ APIs over to Python.