Darshee MacHhar

About Darshee MacHhar

I’m Darshee Machhar, a software engineer working at the intersection of AI/ML and large-scale distributed systems. I currently build foundation model infrastructure at Amazon Web Services (AWS Bedrock), where I work on scaling LLM inference systems that serve billions of requests with low latency and high reliability.A big part of my work is partnering with research teams to take models from experimentation to real-world deployment safely and efficiently.Before this, I built end-to-end ML pipelines in healthcare and worked on reinforcement learning–based NLP systems. I have a Master’s in Computer Science focused on Machine Learning, and I’m especially excited about building intelligent systems that operate reliably at global scale.I have a holistic understanding of computer science principles, scalable solutions, and the data that inform them. I consistently seek opportunities to push technological boundaries and contribute meaningfully to the field. I am eager to collaborate, learn, and grow, ready to make a significant impact in the ever-evolving tech landscape.

Experience

1. Software Development Engineer(Ml)

   Amazon Web Services (AWS)

   Oct 2024 — Present · Seattle, WA, US

   Designed and implemented scalable backend services powering multi-tenant AI model inference, serving millions of daily requests with sub-200ms p99 latency while maintaining high availability at cloud scale.• Developed a high-performance caching layer to handle 200K TPS traffic for inference APIs, caching DynamoDB responses to mitigated hot partition issues, reduce latency and ensure service availability during heavy load.• Engineered a centralized Metadata Service to automate the model onboarding process for Bedrock foundation models, reducing manual intervention and onboarding time by 60%.• Owned CI/CD pipeline health by reducing flaky test failures by 30%, consolidating rollback signals, and introducing granular promotion controls while keeping low end-to-end deployment latency.• Owned and resolved multiple high-severity production incidents by tracing failures end-to-end, closing validation gaps, and implementing fail-fast request validation that eliminated recurring customer-facing 500s.• Designed and implemented request throttling observability improvements by standardizing metrics and enriching telemetry, reducing time-to-diagnose rate-limit issues by 50% across enforcement layers.

Education

• University of Mumbai

  B. Tech., Computer Science

  2016 — 2020

• Udayachal Schools

  SSC

  2008 — 2014

• Rutgers University–New Brunswick

  Master of Science - MS, Computer Science