CV

My curriculum vitae.

Contact Information

Name Yash Vanjani
Professional Title Machine Learning Engineer
Email [email protected]

Professional Summary

Machine Learning Engineer focused on large-scale AI infrastructure, distributed training, and inference systems.

Experience

  • 2026 - present

    Menlo Park, CA

    Machine Learning Engineer
    Meta Superintelligence Labs
  • 2024 - 2026

    San Francisco, CA

    Member of Technical Staff | ML Infrastructure - Pre-training
    Essential AI
    Scaled foundation model pre-training on 1000s of AMD GPUs and Google TPUs.
    • Optimized distributed training for dense models up to 32B, MoE models up to 1T, and context lengths up to 128K
    • Delivered 25% throughput gain and $3M in annual cost avoidance through deep-kernel profiling
    • Innovated layer sharding to improve Muon optimizer performance by 30x for large-scale training
    • Architected distributed inference system in Kubernetes sustaining 1M concurrent requests/s
    • Built KV cache-aware load balancer cutting tail-latency by 40% and saving $2M annually
    • Directed large-scale training dataset curation, raising training efficiency and downstream task accuracy by 10%
  • 2023 - 2024

    San Jose, CA

    Member of Technical Staff 3 | Core Data Path - Stargate
    Nutanix
    • Designed on-demand LSM-memtable checkpointing for rapid crash recovery with near-zero write-path overhead
    • Instituted cross-layer hung-disk detection framework, averting 24-48 hour outages per incident
    • Root-caused and fixed race conditions in the distributed-storage core, cutting fleet-wide crashes by 10%
    • Deployed on-device RAG-powered JIRA bot for auto-summarizing ticket threads
  • 2018 - 2021

    Noida, India

    Software Development Engineer 2 | Adobe FrameMaker
    Adobe
    • Owned and modernized key subsystems: publishing server, online review, update-notification, and logging
    • Hardened middleware design and eliminated deep memory leaks, driving a 100x increase in server availability
    • Architected Document Cloud online review workflows enabling real-time multi-author collaboration
    • Pioneered FrameMaker’s first React-based feature, the Comments Review Panel

Education

  • 2021 - 2022

    Pittsburgh, PA

    Master of Science
    Carnegie Mellon University
    Machine Learning & Systems
    • Coursework: Machine Learning, Deep Learning, Computer Vision, NLP, Deep RL, Search Engines, Distributed Systems, Storage Systems
    • Awarded $10K scholarship from the Information Networking Institute
  • 2014 - 2018

    Guwahati, India

    Bachelor of Technology
    Indian Institute of Technology Guwahati
    Computer Science & Mathematics

Research

Skills

Languages & Frameworks (Expert): Python, C++, JAX, PyTorch, CUDA, Triton, Pallas
ML Infrastructure (Expert): vLLM, SGLang, MaxText, Distributed Training, Model Parallelism
Infrastructure & Tools (Expert): Kubernetes, Spark, Grafana, Airflow, GCP, Terraform

Awards

  • 2021
    $10K Graduate Scholarship
    Information Networking Institute, CMU

    Awarded for graduate studies at Carnegie Mellon University.

  • 2014
    All India Rank 1875 in IIT-JEE Advanced
    IIT-JEE

    Secured among 1.4 million participants.

Invited Talks

Interests

Large-scale Distributed Training: Data Parallelism, Tensor Parallelism, Pipeline Parallelism, FSDP, MoE Training
LLM Inference: KV Caching, Speculative Decoding, Continuous Batching, Prefix Caching
ML Systems: GPU Kernel Optimization, Distributed Systems, Data Pipelines