Oct 2019 – Jan 2020 · 3 mos · London, United Kingdom

• Launched Revolut's first regulatory compliant crypto product in the US.

Jan 2019 – Oct 2019 · 9 mos · London, United Kingdom

• Reduced payment fraud rates during period of explosive customer growth.

Apr 2016 – Jul 2020 · 4 yrs 3 mos · San Francisco Bay Area

• Risk Salon is a think tank that brings together people working in the broad risk space, including payments risk, trust and safety, and compliance, at tech companies in the SF bay area and Seattle. We organize monthly roundtable forums and an annual risk conference called Deflect.

Jul 2015 – Jan 2019 · 3 yrs 6 mos · San Francisco Bay Area

• Coinbase is the #1 cryptocurrency Exchange in the US. I was responsible for various critical functions including, Data, Fraud, User Security, Identity & Internal Tools, comprising a team of 40 engineers, data scientists, and managers. Some of my accomplishments include:
• RISK:
• Reduced payment fraud rate by 10x and maintained it during high-growth
• Stopped several account takeover threats (SIM swap, social engineering)
• Built high-limit card and ACH purchase flows including bank verification
• Launched enhanced user security features: (i) account recovery on lost phone/email (ii) enhanced 2FA (SMS, TOTP) (iii) delay suspicious logins or withdrawals and accelerate it on selfie upload
• Launched KYC: (i) selfie in KYC flow (ii) deep learning based Facematch to match fraud selfies (iii) liveness detection using machine learning
• DATA:
• In 2015, I wrote our first machine learning training pipelines based on Airflow & Vowpal Wabbit, and later Amazon Sagemaker.
• Built a robust ETL system to load production data into Redshift and later Snowflake
• Built a streaming system to perform in-stream computation on log data using Kinesis & Apache Flink
• Built our AB testing infrastructure to configure AB tests and monitor metrics automatically
• Launched several machine learning models to detect fraud & compliance bad actors
• Deep learning models to extract faces from driver licenses and build tamper-proof ID similarity
• Blockchain analytics infrastructure to streamline our accounting and track movement of crypto by bad actors

Mar 2014 – Jun 2015 · 1 yr 3 mos · San Francisco Bay Area

• Built an ad-targeting engine that matched search queries to relevant ads using Click Through Rate (CTR) prediction. Also worked on Yelp's auto-bidder to help local advertisers manage their bid prices based on demand and budget constraints.

Apr 2013 – Mar 2014 · 11 mos · San Francisco Bay Area

• First Data Science Engineer at Flurry, responsible for building out our Machine-Learning based Real-Time Bidder (RTB). This ML-based strategy would determine which mobile impressions to bid on and the bid-value.
• ML models to predict Click-Through Rates (CTR) and Conversion Rates led to 10-30% CPM lifts.
• Developed an automated ML pipeline using Map-Reduce, Hbase, Java, Liblinear
• Built a recommender system to predict which apps a user is likely to install

Apr 2011 – Mar 2013 · 1 yr 11 mos · San Francisco Bay Area

• MyLikes is a social-media ad network that connects advertisers with audience via influencers. Influencers are paid per click for promoting a product in their feed on Twitter, Facebook, YouTube.
• Fraud systems: to detect click-fraud, phishing ads, and spammy publishers
• Pricing algorithms to determine publisher’s Cost Per Click (CPC) based on click quality

Nov 2005 – Apr 2011 · 5 yrs 5 mos

• Worked on a team of 6 engineers that built a cyber-security product for large telcos to detect and mitigate bad traffic: Botnets, DDoS attacks, 0-day worms, DNS Cache Poisoning, BGP routing anomalies, Call-spam.
• Other things I did in my tenure there:
• Managed an academic research program with $1M budget with 10 top-tier CS universities
• Recruited and managed 20 Ph.D. interns and 10 CS research faculty
• Co-authored 18 patents and 25 research papers (KDD, SIGCOMM, SIGMETRICS, etc)
• Helped enterprise sales team close multi-million dollar deals at various ISPs over the world

Jul 2003 – Nov 2003 · 4 mos

• Engineering intern on Search infrastructure team at Google. Built a bloom-filter like approach in the search-index balancer to cleverly route queries to index-servers most likely to have the corresponding reverse-indices.

Jun 2001 – Sep 2001 · 3 mos

• Built a demand-based resource provisioning system for managing web clusters

Aug 2000 – Aug 2005 · 5 yrs