Skip to content
Resume

Bangalore, India/IST

Hi 👋 I'm Arka Pramanik

Software Engineer @ HPE

I build Kubernetes platform services and LLM agents for the software that runs HPE's Cray supercomputers.

View my workResume
rack-failure.sim
Healthy
monitor
pending 0
Services up
6/6
Replicas
18/18
Availability
100%

› All zones healthy. Click a rack to cut its power, or a pod to kill it.

A toy model of the failures my Rack Resiliency Service work guards against.

01 / About

Introducing Myself

Hi, I'm Arka, a Software Engineer at Hewlett Packard Enterprise in Bangalore. I work on Cray System Management, the software stack that runs HPE's Cray supercomputers.

My work sits where Kubernetes platform engineering meets applied AI. I conceived and lead ATLAS, an AI root cause analysis tool that cut manual triage by about 80%, and co-developed the Rack Resiliency Service, which keeps 59 critical services highly available when a rack fails.

Before HPE I maintained an Android device tree for ArrowOS and spent a lot of evenings on competitive programming. I hold the CKAD and Azure Fundamentals certifications, and I mentor in the HPE CTY program.

whoami.yaml
critical services kept highly available by RRS
59
less manual triage with ATLAS
~80%
MCP tools powering LLM agents
15
Jira tickets in semantic search
3,291

02 / Journey

My Experience

  1. Hewlett Packard Enterprise

    Systems/Software Engineer 1, HPC Cray System Management

    Sep 2024 – PresentBangalore, India

    ATLAS

    Automated Triaging and Log Analysis Solution, an internal AI root cause analysis tool

    • Conceived and now lead ATLAS, a 3-stage AI pipeline (50+ health checks, FAISS similarity search, LLM root cause analysis) that cut manual triage by about 80%.
    • Architected the Root Cause Analyzer as a Python MCP server whose 15 tools let GitHub Copilot agents gather Jira, GitHub and documentation evidence and write RCA reports autonomously.
    • Rolled out ATLAS to 3+ teams as an official HPE tool, delivering RCA reports in about 2 minutes.

    Rack Resiliency Service

    Rack-level high availability for the Cray System Management stack

    • Co-developed RRS end to end for CSM 1.7, providing rack-level high availability for 59 critical services.
    • Implemented the Flask REST API (7 endpoints, OpenAPI 3.1) and RRS CLI (7 commands), letting operators inspect rack zones and register critical services through JSON payloads.
    • Built the monitoring service that detects node or rack failures and flags imbalanced services within 10 minutes.
    • Enforced rack-aware replica placement with a Kyverno policy, validated on 3- and 4-zone bare-metal layouts.
    • Packaged RRS as 4 Docker images with Jenkins CI/CD and made 4 Ansible roles idempotent.

    Platform Engineering

    Service mesh, upgrades and day-to-day ownership

    • Delivered the Istio 1.19.10 to 1.26.0 upgrade for Kubernetes 1.32, replacing the operator with Helm charts.
    • Owned platform and hardware-management services, closing 10 triage issues and 2 customer cases.
    • Fixed etcd health checks that passed with 0 ready pods and automated CRD upgrades via Helm hooks.
    • Python
    • Kubernetes
    • Flask
    • FastAPI
    • Model Context Protocol
    • FAISS
    • Helm
    • Istio
    • Docker
    • Kyverno
    • Ansible
    • Jenkins
    • Bash
    • Linux
  2. Hewlett Packard Enterprise

    HPC Software Engineer Intern

    Feb 2024 – Aug 2024Bangalore, India

    • Upgraded Istio from 1.11.8 to 1.19.10 (8 minor versions), enabling Kubernetes 1.24.
    • Validated the upgrade through 5 POCs spanning Istio 1.13 to 1.19.
    • Rebased HPE's Istio fork on upstream, updated 3 Helm charts and upgraded Kiali to 1.75.0.
    • Kubernetes
    • Istio
    • Kiali
    • Helm
    • Docker
    • Git
    • Linux
    • Bash
  3. Hewlett Packard Enterprise

    HPE CTY Summer Intern

    Jan 2023 – Jun 2023Remote

    • Created a Grafana security dashboard for Kubernetes covering 116 CIS checks and image vulnerabilities.
    • Automated hourly Trivy scans with Helm, 2 scripts, Prometheus and Node.js.
    • Kubernetes
    • Grafana
    • Prometheus
    • Trivy
    • Helm
    • Node.js
    • Docker
    • Bash
  4. ArrowOS

    Device Maintainer (open source)

    Feb 2021 – Aug 2022Remote

    • Maintained the Asus Zenfone Max M2 (SDM632) device tree for ArrowOS on Android 12.
    • Improved the day-to-day experience through device-side fixes and performance work.
    • Worked with the open-source community on custom Android ROM builds and releases.
    • Android
    • Linux Kernel
    • C/C++
    • Makefile
    • Bash
    • Git

Prefer the one-page version?

Download resume

03 / Projects

My Recent Works

Here are a few projects I've worked on recently.

Internal HPE toolfeatured / 01

ATLAS Framework

Automated Triaging and Log Analysis Solution

An AI root cause analysis tool for HPC clusters. Hand it a Jira ticket or an error log and LLM agents gather the evidence and write a root-cause report with remediation steps.

  • Engineered agent context via tiered retrieval and 3-layer de-duplication so the LLM reads all evidence at once.
  • Shipped semantic search with BGE-M3 embeddings and re-ranking across 3,291 Jira tickets.
  • Launched the web UI and REST API with FastAPI and GitHub Copilot SDK, backed by 270+ pytest tests.
  • Python
  • Model Context Protocol
  • FastAPI
  • FAISS
  • RAG
  • GitHub Copilot SDK

Source stays inside HPE.

pipeline

  1. Jira ticket or error logPods restart in a loop after last night's upgrade.
  2. 01Health checks50+ automated checks
  3. 02Similarity searchFAISS over past tickets
  4. 03LLM root cause analysisagents with 15 MCP tools
  5. RCA reportin about 2 minutes
Root cause
The new chart renamed a config key that the service still reads at startup.
Fix
Add the new key to the values file and roll the deployment.

A simulated run on made-up incidents. The real pipeline takes about 2 minutes.

~80%
less manual triage
15
MCP tools
~2 min
per RCA report
Open sourcefeatured / 02

Rack Resiliency Service

Rack-level high availability for Cray System Management

A Kubernetes service that keeps critical services running when an entire rack goes down, by placing replicas across rack zones and watching for imbalance.

  • Implemented the Flask REST API (7 endpoints, OpenAPI 3.1) and RRS CLI (7 commands) for operators.
  • Built the monitoring service that detects node or rack failures and flags imbalanced services within 10 minutes.
  • Enforced rack-aware replica placement with a Kyverno policy, validated on 3- and 4-zone bare-metal layouts.
  • Python
  • Flask
  • Kubernetes
  • Ceph
  • Helm
  • Kyverno
  • Ansible
View source
59
critical services
7 + 7
API endpoints + CLI commands
CSM 1.7
shipped in
  • Screenshot of HPC Security DashboardWork

    HPC Security Dashboard

    A Grafana dashboard that tracks 116 CIS benchmark checks and container image vulnerabilities for Kubernetes, refreshed hourly from Trivy scans.

    • Kubernetes
    • Grafana
    • Prometheus
    • Trivy
    • Helm
    • Node.js
    Code
  • apxd.in
    Open source

    Personal Portfolio

    This site. A static, hand-built portfolio with a live rack-failure simulation, a command palette and a terminal, in under 20 kB of JavaScript.

    • Astro
    • TypeScript
    • Tailwind CSS
    CodeLive
  • Screenshot of ArrowOS Device TreeOpen source

    ArrowOS Device Tree

    Device tree for the Asus Zenfone Max M2 (X01AD), used to build ArrowOS custom ROMs across several Android versions.

    • Android
    • Linux Kernel
    • C/C++
    • Makefile
    Code
  • Screenshot of Online Tuition Management SystemAcademic

    Online Tuition Management System

    A web app for local tuition centres: teachers upload marks, attendance and course material, and students and parents track progress from their own portals.

    • Django
    • Python
    • SQLite
    • Selenium
    Code
  • Screenshot of HabifyAcademic

    Habify

    A habit tracker app that helps people stay disciplined and build good habits over time.

    • React Native
    • Firebase
    • JavaScript
    Code
  • Screenshot of EnConnectAcademic

    EnConnect

    A private chat app that encrypts and decrypts every message with several standard algorithms chained in one pass.

    • React
    • Node.js
    • Socket.IO
    • Encryption
    Code
  • Screenshot of YouTube AnalyserAcademic

    YouTube Analyser

    Produces a crisp text summary of a YouTube video using natural language processing.

    • Python
    • Flask
    • NLP
    Code

04 / Skills

My Tech Stack

The tools I reach for most, grouped by where they show up in my work.

Languages & Frameworks

~/languages/
  • PythonPrimary language
  • FlaskWeb framework
  • FastAPIAsync APIs
  • TypeScriptTyped JavaScript
  • JavaScriptWeb & Node.js
  • GoSystems language
  • BashShell scripting
  • C/C++Competitive programming
  • SQLDatabases

Cloud & DevOps

~/cloud-devops/
  • KubernetesOrchestration · CKAD
  • DockerContainers
  • HelmPackaging
  • IstioService mesh
  • AnsibleConfiguration
  • JenkinsCI/CD pipelines
  • GitHub ActionsCI/CD pipelines
  • CI/CDDelivery automation
  • LinuxOperating system

Backend & AI

~/backend-ai/
  • REST APIsOpenAPI 3.1
  • MicroservicesService design
  • Generative AILLM applications
  • Agentic AITool-using agents
  • RAGRetrieval pipelines
  • Model Context ProtocolAgent tooling

Tools & Testing

~/tools/
  • GitVersion control
  • JiraIssue tracking
  • PrometheusMonitoring
  • GrafanaDashboards
  • pytestTesting

My Coding Profiles

Where I practise data structures, algorithms and contest problem solving.

My Certifications

My Achievements

  • Mentor, HPE CTY Program 2026Back in the program that started it all, this time as a mentor.
  • Top 25, HPE CTY Program 2023Selected among the top 25 from the university.
  • Top 200 in IndiaAmazon ML Summer School 2022.
  • 96.67 percentileJEE Main.
  • GirlScript Summer of Code 2022Open-source contributor.
  • AIR 1403Google Hash Code 2021.

05 / Contact

Connect With Me

Let's Connect,

Always excited to connect with like-minded professionals and collaborate on innovative projects.

Drop me a message at

Bangalore, Indiamy time: IST

Ready to build something impactful?

Goes straight to my inbox.