Datta Vellal

2018

Took on cost and reliability for the fraud services around Prime Day: cut garbage collection on one service to about 2.4% of CPU, saved about $36K a year in hosts through load tests and a hardware clean-up, and traced a faulty fraud rule on Prime Day and escalated it as a sev-2. Became the trainer program's highest-scoring trainer, then left in August to join Philips.

  • ~17.7%to ~2.4% GC CPU
  • ~60%less investigation effort
  • 5sessions

~$36K a year saved in hosts. 4.57/5 top trainer.

What mattered most

  1. 01

    Highest-scoring trainer after teaching about 90 people Scrum

    My last Scrum workshop at Amazon scored 4.57 out of 5 and made me the program's highest-scoring trainer that month, after about 90 people across 3 Scrum workshops.

  2. 02

    Caught an auto-approval bug in a fraud rule on Prime Day

    Under Prime Day volumes, ruleset failures were letting wallet top-up orders pass without a fraud check; I traced them to one wrongly written rule, pointed to the exact coding error and, after checking with my manager, escalated a sev-2.

  3. 03

    Sized two fraud services for Prime Day from load tests, $12,344 a year saved

    When the org scaled up for Prime Day, I load-tested two fraud services, set each host count from the results and cut 18 hosts.

  4. 04

    Cut investigation effort about 60% on the Amazon Pay latency project

    While the org moved Amazon Pay top-up fraud checks to a new engine to cut latency, I reworked the tool that compares old and new fraud variables so engineers on two teams and the ML scientists could find and fix mismatches fast.

  5. 05

    Cut garbage collection to 2.4% of CPU on a fraud service

    I profiled a fraud-prevention service, found it spent about 17.7% of CPU on garbage collection, tuned the JVM to bring that to 2.4%, and showed the team the before-and-after evidence.

“You have shown great example of Deep Dive. Finding out exactly which rule was incorrect and also pointing to exact coding error.”

Engineering managerAmazon, July 2018Read the capsule

By month

Also in 2018

  1. August 2018Left Amazon for Philips after about 30 months, 68 interviews and 570 code changesRole
  2. 15 March 20182018 annual review: my manager named hiring and developing people, process and documentation as superpowersReview
  3. March 2018Called in to fix critical incidents, including a sev-2 spike of 6,400 manual investigations in three hoursProject
  4. 16 February 2018Taught my first Agile Product Ownership workshop to a full class of 30 and took the pace feedback on boardTalk
  5. January 2018Owned a hardware clean-up that moved fraud services to new-generation hosts and saved about $23,800 a yearProject
  6. 2018Reviewed 3 promotion documents for engineers moving to SDE II and mentored interns and junior engineersHiring
  7. 2018Taught 5 sessions in 2018 to about 230 people across Amazon and PhilipsTalk