← Back to home

HPC Development Engineer · 9+ years R&D

Gahan Saraiya

High Performance Computing (HPC) Development Engineer with 9+ years of Research and Development experience in system software, GPU computing, and distributed systems architecture. Currently at AMD India Pvt Ltd., specializing in performance optimization, AI/ML workload execution on HPC accelerators (GPUs), and cluster observability (Prometheus, Grafana, Slurm). Deep expertise in Python, modern C++, and Bash for large-scale on-prem and cloud deployments.

Email LinkedIn GitHub Resume (R&D PDF)

Current role

Senior System Software Designer

AMD India Pvt Ltd · Mar 2024 – Present
  • Cluster Observability: Prometheus, Grafana, Slurm telemetry; auto-deploy observability runners and agents for large GPU fleets.
  • Agentic Memory Network: Obsidian design for agent memory index, knowledge base, and skills auto-sync (human-readable and agent-readable).
  • Telemetry Pipeline: Secure log upload and cluster validation telemetry shipping to AWS S3.
  • R&D Leadership: Next-generation AMD GPU accelerators (MI2XX, MI3XX series).
  • Performance Engineering: GPU benchmarking with modern C++ and Python.
  • Deep Learning: PyTorch on large-scale GPU clusters — GEMM and distributed training pipelines.
  • HIP/ROCm: Compute partitioning, HIP kernel profiling, RAS injection, parallel execution.

Technical skills

Languages C++, C, Python, Bash, SQL
HPC & GPU ROCm, HIP, CUDA, MPI, OpenMP, GPU profiling
Deep Learning TensorFlow, PyTorch, TensorRT, ONNX, distributed training
Observability Prometheus, Grafana, Slurm, telemetry, S3 log shipping
Agentic AI Obsidian KB, LangGraph, FAISS, OpenAI, Vertex AI
Infrastructure AWS, GCP, Azure, K8s, Docker, Terraform, Jenkins
Systems Linux internals, UEFI/BIOS, RAS, SMI, CPER

Projects

2026 – Present

Cluster Management and Observability

Prometheus, Grafana, and Slurm telemetry; auto-deploy observability runners and agents; secure log shipping to AWS S3 for cluster validation CI.

2026 – Present

Obsidian Agentic Memory Network

Agent memory index, knowledge base, and skills auto-sync for human-readable and agent-readable consumption.

2025 – Present

Agentic AI for Software Engineering

Multi-tool skills, agents, and hooks (Claude, Cursor) with security compliance and token spend restrictions.

2025 – Present

Context-Aware Task Routing with LangGraph

LangGraph routing with FAISS semantic search; GitHub/Jira connectors; OpenAI, Vertex AI, and local LLMs.

Defect triage RAG notes

2022 – 2024

Universal Scalable Firmware (USF)

Platform orchestration layer for firmware configuration; UEFI industry contribution.

UEFI DevCon presentation

2020 – Present

Firmware Automation Framework

SDL-compliant edk2 UEFI automation — Jenkins, GitHub Actions, JFrog Artifactory.

intel/xml-cli

Achievements

Publications & talks

Certifications

Education