Revant Gupta

About Revant Gupta

Software Engineer with 5+ years of experience building scalable backend systems and ML infrastructure using Rust, Python, and AWS. I design for performance, reliability, and cost-efficiency across edge and cloud. At Amazon Go, I led the shift to ARM-based edge inference—cutting compute costs by 80% and scaling to 200+ stores. Designed deployment frameworks adopted by 20+ teams and accelerated model launch cycles by a month by streamlining accuracy validation. Improved ML pipeline throughput by 20%, halved recovery time (5→2.5 days), and enabled support for temporal models—reducing ECA by 41% and saving \\$300/store/year. Built GenAI tools using LangGraph and Bedrock to automate monitoring workflows, saving 4+ dev hours/week. I drive architecture, cross-team enablement, and measurable system improvements at scale.

Experience

1. Software Development Engineer 1, 2

   Amazon

   May 2021 — Present · Bangalore Urban, IN

   Developed a Rust-based Android application that optimized batch processing of RGB datasets, enabling efficient downloading from S3, SSE decryption, and decoding of frames. This system synchronized spooling of RGB frames at a constant 15 FPS to over 130 on-premise cameras, significantly enhancing the accuracy and performance testing of ML models across diverse RGB datasets- Pioneered organization\'s first ARM-compatible service on specialized edge devices, resolving critical scaling bottlenecks and achieving 80% cost reduction by eliminating cloud infrastructure, successfully piloting in 2 stores before scaling to 200+ stores in both cloud and on-premise environment- Engineered hybrid infrastructure framework supporting both edge and cloud deployments, enabling 20 teams to leverage shared computing devices while maintaining cloud failover capabilities through unified CloudFormation templates, deployment pipelines, and cross-environment monitoring solutions via CloudWatch- Architected and implemented offline support for DART-X unified tracking system, engineering sophisticated post-processing pipeline for variable frame rates (5-15 FPS) and resolutions (VGA/qVGA), reducing inference latency by 67% while enabling 41% ECA reduction per rack and $0.3K/store/year CapEx savings- Developed parametrized configuration framework with innovative frame-overlapping solution for temporal models, accelerating research iteration cycles for 16+ model versions without code modifications while maintaining backward compatibility with production systems.Tools: Rust, Java, Python, C++, AWS (ECS, SQS, S3, CloudWatch, EC2, KVS, AppConfig)

Education

• Scaler

  Computer Science

  2019 — 2021

• KRISHNA INSTITUTE OF ENGINEERING AND TECHNOLOGY, GHAZIABAD

  Bachelor of Technology - BTech

  2016 — 2020

• Delhi Public School, Gurgaon (Sector 45)

  Physics, Chemistry, Maths, C++

  2012 — 2016