about

Software engineer at Meta, working on large-scale P2P content distribution with the Core Systems group. Previously at AWS Fraud Prevention for 4 years — building distributed systems to protect customers from credential compromise and account takeovers across all AWS regions.

MS Computer Science from UT Dallas. Interests outside work: certified Rescue Diver, open-world games, reading.

viraj.mavani1996@gmail.com  ·  github  ·  linkedin  ·  scholar  ·  @vdmavani

writing

• 2026-02-25
  The Engineering Behind Serving LLMs at Scale

  software
• 2023-09-12
  The Framework Dilemma: Are Frameworks the Death of Software Engineering?

  software
• 2023-05-14
  Containerization in Distributed Systems

  software
• 2023-04-16
  Cost of Building Monolithic Applications

  software
• 2020-05-25
  Product of Array Except Self — LeetCode #238

  leetcode
• 2020-04-22
  Contains Duplicate — LeetCode #217

  leetcode
• 2020-04-20
  Two Sum — LeetCode #1

  leetcode

experience

• Meta 2025-05 – now
  Software Engineer, Core Systems
  Large-scale P2P content distribution.
• Amazon Web Services 2021-06 – 2025-05
  Software Development Engineer, Fraud Prevention
  Distributed systems for fraud detection across all AWS regions.
• Walmart Inc. 2020-06 – 2021-05
  Software Engineer II, Fraud Detection
  Fraud Detection Engine preventing $100M/yr in losses. Python, Spark, Databricks.
• 7-Eleven Inc. 2019-05 – 2019-08
  Machine Learning Intern
  Time-series forecasting with Facebook Prophet. 5.36% mean error. React + Node.js dashboard.
• SeRViCE Lab, UT Dallas 2018-09 – 2019-05
  Research Assistant
  Visual SLAM loop-closure detection for autonomous vehicles. PyTorch.
• IIT Gandhinagar 2017-05 – 2018-06
  Research Intern
  Facial expression recognition via deep learning. Published at ICCV 2017, CVPR 2018.

skills

projects

• AnnoMage
  Semi-automatic image annotation tool using RetinaNet (PyTorch). CSV and Pascal VOC XML output.
• Ensemble Decision Trees
  Bagging and AdaBoost on ID3 trees — pure Python and NumPy, no ML libraries.
• DavisBase
  RDBMS from scratch in C++17. B+Tree indexing, persistent storage, SQL-like interface.
• Bifrost
  Browser new-tab extension with search filters. Serverless Spring Boot API backend.
• Raster Map Intersection Detection
  Real-time road intersection detection from raster maps + Google StreetView classifier.

publications

1. SAF-BAGE: Salient Approach for Facial Soft-Biometric Classification — Age, Gender, and Facial Expression
   A Gurnani, K Shah, V Gajjar, V Mavani, Y Khandhediya · WACV 2019
2. ViSHuD: Using Visual Saliency to Improve Human Detection with CNNs
   V Gajjar, Y Khandhediya, A Gurnani, V Mavani, M Raval · CVPR 2018
3. Facial Expression Recognition using Visual Saliency and Deep Learning
   V Mavani, K Miyapuram, S Raman · ICCV 2017

reading