Skip to content

saatwik@dev:~$ whoami

saatwik yajaman

I build the systems that stay up, and the apps that run on them.

saatwik@dev:~$ help

try ls, cat status.txt, cat meta, open everygpu, contact. tab completes, up arrow recalls.

work

  1. CybOrg at USC

    Infrastructure Tech Lead, TACTICS

    Sep 2026 - Present, Los Angeles, CA

    • Executed a platform migration across GCP environments to sustain weekly CTF events, managing Kubernetes infrastructure and Traefik ingress routing to dynamically spin up isolated challenge containers for 20+ competitors.
    20+
    Competitors per weekly CTF
    • GCP
    • Kubernetes
    • Traefik
  2. Meta

    Production Engineer Intern

    May 2026 - Aug 2026, Bellevue, WA

    • Streamlined six-week drain test scheduling across Meta's products and services, cutting planning time from 1 day to 2 hours, by deploying a Python LLM agent adopted by the Disaster Recovery (DR) team.
    • Preempted service failures during drain tests by publishing daily forecasts of at-risk services and failure modes, giving service owners a 7-day advance notification window.
    • Quantified drain test ROI across 15+ historical tests for the DR team by designing an evaluation framework that correlated SEV root causes with risk telemetry to isolate novel failure modes.
    Drain test planning time: 1 day to 2 hours
    7 days
    Advance warning for service owners
    • Python
    • LLM agents
    • Risk telemetry
  3. Crio.Do

    Product Engineer Intern

    Jan 2025 - Jul 2025, Bengaluru, India

    • Delivered an end-to-end AWS deployment curriculum for 50+ learners by building starter stubs, reference solutions and Docker Compose setups across EC2 and Lambda for a four-microservice app.
    50+
    Learners taught to deploy on AWS
    4
    Microservices in the reference app
    • AWS EC2
    • AWS Lambda
    • Docker Compose
  4. ANB Solutions

    Software Engineering Intern

    Aug 2024 - Dec 2024, Remote

    • Automated e-commerce product cataloging, reducing manual tagging effort by 50%, by deploying a multimodal pipeline using a Llama 3 vision model to extract and validate structured metadata from item images.
    Manual tagging effort: 100% to 50%
    • Llama 3 Vision
    • Python
    • Multimodal pipelines
  5. Google Summer of Code

    Software Contributor @ Catrobat

    Jun 2022 - Oct 2022, Remote

    • Enhanced drawing precision across 13 canvas tools, shipped to production for 1M+ users, by implementing an interactive canvas magnifier with dynamic coordinate mapping and viewport positioning.
    • Prevented UI regressions across an Android app with 1M+ downloads by building reusable UI test helpers and 4 Espresso E2E tests validating visibility and placement.
    1M+
    Users reached in production
    13
    Canvas tools made more precise
    • Kotlin
    • Android
    • Espresso

Contact

Let's build something.

Always up for a conversation about systems, infrastructure, and the products built on them.

saatwik.sy@gmail.com