Ayush Panigrahi — AI Platform Engineer

01 / Introduction

The architect behind the systems.

Designing systems that endure.

Over the past 1.5 years, I've architected company-wide API key management, led UnifyAI 2.0 platform architecture, and engineered production-grade AuthZ/RBAC systems using Keycloak, OpenFGA, and APISIX.

I think in systems — ETL pipelines, multi-agent orchestration, K8s operators, encryption layers. When something breaks at 2am in a client deployment, I'm the one picking up the phone.

0

Experience

0

GPA / 10

0

Mentored

Research

ICAISI 2025 — GenAI collaboration platforms w/ personalized RAG

Achievement

Winner, Hackscript 2025 · IEEE SB Chairperson

Optimization

LLaMA 3.2 & Qwen3 on AMD MI300 — 25% memory cut, <10s inference

02 / Career Path

Roles & responsibilities

Aug 2025 — Present

AI Platform Engineer

SK International, Dadar, Mumbai

Architected company-wide API key management infrastructure with scope-based access control, public/private key lifecycle management, and audit logging.
Led UnifyAI 2.0 platform architecture: integrated Apache NiFi, Airflow, and OpenMetadata for automated ETL pipelines and data governance.
Engineered production-grade AuthZ/RBAC using Keycloak, OpenFGA, and APISIX — fine-grained user isolation across Agentic AI and LLM modules.
Led API gateway migration from Kong to APISIX, optimizing routing, rate-limiting, and security policies for LLM endpoints.
Built custom encryption layer for data-in-transit between Airbyte and PostgreSQL; deployments passed rigorous enterprise security audits.
Primary technical point of contact for clients — business requirements → engineering features → post-deployment support.

Feb 2025 — Jul 2025

MLOps / DevOps Intern

Data Science Wizards

Deployed RAG agents on Kubernetes using Helm and HPA; reduced query latency by 30%.
Engineered Kafka + CDC data aggregator pipelines on K8s, cutting data sync time by 40%.
Orchestrated Apache Airflow on Kubernetes with custom StatefulSets and PVs for versioned ML workflows.

03 / Featured Work

Builds & deployments

Smaran

Python · MCP · LLM Agents

AI agent memory management system exposed via MCP protocol. Installable Python package enabling persistent structured memory across production Agentic AI workflows.

GitHub ↗

Shell Story

Python · Swarm Agents · Markdown/PDF

Terminal session recorder feeding shell commands to a swarm of agents. Auto-generates structured deployment documentation for future agent training. In progress.

GitHub ↗

k8s-mcp & ks-ai

Go · Kubernetes · Gemini API

Go-based multi-cluster pod management CLI with TUI. Natural language Kubernetes operations via Gemini-powered agents. Talk to your clusters like coworkers.

GitHub ↗

PruneX

Python · SHAP · PyPI

End-to-end automated feature engineering library. Preprocessing, encoding, dimensionality reduction, and leakage detection using deterministic statistical methods.

PyPI ↗ GitHub ↗

05 / Elsewhere