Lokesh Kumawat.

Software Engineer · Jaipur, Rajasthan, India

Building enterprise GenAI & agentic AI systems.

Hi, I'm Lokesh Kumawat — a software engineer at Celebal Technologies building enterprise-grade GenAI and AI-agent platforms on Microsoft Azure. I work across backend services, agent orchestration, and enterprise integration, turning complex requirements into reliable, secure, well-architected systems.

  • Python · FastAPI
  • Azure AI Foundry & OpenAI
  • Agentic AI · RAG
  • AKS & Kubernetes
  • Cosmos DB
  • Enterprise integration
Read the articles →Get in touch
Lokesh Kumawat — Software Engineer

About

My journey has moved from strong engineering foundations to production platform work. I started by mastering Python and backend development — building REST APIs with FastAPI (Pydantic, JWT auth, middleware), working with databases through SQLAlchemy, and getting hands-on with Azure (AZ-900 certified): Functions, Logic Apps, Cosmos DB, Event Grid, and Event Hub. I then expanded into cloud-platform and DevOps engineering — Docker, Kubernetes/AKS, Helm, KEDA, Azure API Management, Azure DevOps CI/CD, and infrastructure with ARM templates.

Today I work on AI-agent orchestration and enterprise integration: developing backend services that connect LLM-powered agents to enterprise systems (Microsoft 365, SAP, ServiceNow, Databricks), and applying core GenAI concepts — RAG, prompt engineering, tool/function calling, and structured outputs — using Azure AI Foundry and Azure OpenAI. My work spans client and internal projects covering workflow-automation engines, human-in-the-loop systems, governance/observability, agent-framework migration, and performance/architecture optimization.

I enjoy turning complex requirements into reliable, secure, well-architected systems — and I'm focused on growing as a backend and AI platform engineer building real-world agentic AI.

Core stack — Python · FastAPI · Azure AI Foundry · Azure OpenAI · Cosmos DB · Azure Functions · AKS · Azure API Management · SQLAlchemy · Databricks · RAG · Agentic AI

Experience

  1. CT

    Celebal Technologies

    1 yr 8 mos · Jaipur, Rajasthan, India · On-site

    1. Software Engineer

      Full-time

      Jul 2025 - Present · 1 yr

      I build and ship the backend behind our enterprise GenAI products — designing APIs and services in Python and FastAPI and integrating them with Azure AI Foundry and Azure OpenAI. Much of my work is on Agent Garage, Celebal's platform for building and governing AI agents at scale, where I help agents reliably orchestrate multi-step workflows and connect securely to enterprise systems like Microsoft 365, SAP, and ServiceNow. It's taught me what it really takes to make AI dependable in production — handling real-world edge cases, failures, and scale, not just the happy path in a demo.

    2. Jr. Associate

      Apprenticeship

      Mar 2025 - Jun 2025 · 4 mos

      This is where I went deep on cloud platform and DevOps. I learned to run containerized workloads properly on Kubernetes/AKS — deployments, ingress, Helm, and autoscaling — and set up CI/CD with Azure DevOps and infrastructure as code with ARM templates. I also explored building MCP servers on Azure Functions, and got confident enough with the platform to run internal knowledge-sharing sessions on Kubernetes and AKS for the team.

    3. Jr. Associate

      Internship

      Nov 2024 - Feb 2025 · 4 mos

      My software-engineering foundations were built here. I got hands-on with Python and wrote my first real REST APIs in FastAPI — learning request validation, authentication, and middleware properly — and worked with databases through SQLAlchemy. I also earned my first Azure certification (AZ-900) while getting practical experience with Azure Functions, Cosmos DB, and enterprise auth patterns like OAuth 2.0 and SSO.

  2. CT

    Data Engineering Intern (COE Program)

    Celebal Technologies · Remote · Internship

    May 2024 - Aug 2024 · 4 mos

    My first taste of working with data at scale. Through Celebal's COE program I learned the fundamentals of data engineering — building pipelines and transformations with Python, SQL, PySpark, and Databricks — and worked on real projects that taught me how to handle and optimize data for actual business use cases.

Projects

Agent Garage

Enterprise Agentic Process Automation platform · Celebal Technologies

Agent Garage is Celebal Technologies' enterprise Agentic Process Automation platform for designing, deploying, and governing AI agents at scale. As part of the team, I work on the workflow-orchestration and enterprise-integration capabilities — enabling no-code automation of complex, multi-step business processes with secure connectivity to enterprise systems (Microsoft 365, SAP, ServiceNow, Databricks) and built-in governance and observability.

  • Agentic AI
  • Workflow orchestration
  • Enterprise integration
  • Azure

Skills & Tools

Languages
Python, SQL
Backend & APIs
FastAPI, REST APIs, SQLAlchemy, Pydantic, OAuth2 / OpenID / JWT
GenAI & Agentic
Azure AI Foundry, Azure OpenAI, RAG, Prompt engineering, Tool / function calling, Agentic AI, MCP
Cloud & DevOps (Azure)
Azure Functions, Logic Apps, Cosmos DB, Event Grid / Hub, AKS / Kubernetes, Docker, Helm / KEDA, Azure API Management, Azure DevOps CI/CD, ARM templates
Data
Databricks, PySpark, Azure Data Factory

Education

  1. Poornima College of Engineering

    B.Tech, Computer Science

    2021 – 2025

    Grade: 7.93 CGPA

  2. Warren Academy School

    2011 – 2020

    Class 4th – 12th

Certifications

  1. Claude Certified Architect – Foundations

    Anthropic

    Issued Apr 2026

  2. Databricks Certified Data Engineer Associate

    Databricks

    Issued Jan 2026

  3. Microsoft Certified: Azure Developer Associate · AZ-204

    Microsoft

    Issued Jun 2025

  4. Microsoft Certified: Azure Fundamentals · AZ-900

    Microsoft

    Issued Jun 2025

Blogs

All posts →

June 16, 2026

Designing idempotent APIs

Networks retry, clients double-click, and queues redeliver. Here is how to make sure none of that charges a customer twice.

APIs · Reliability

May 28, 2026

Message queues without the foot-guns

Queues decouple your services beautifully — right up until a poison message takes down a consumer at 2am. A field guide.

Messaging · Distributed Systems

May 9, 2026

Observability: start with three signals

You don't need a full observability platform on day one. You need to answer 'is it down, is it slow, and why' — in that order.

Observability · Operations

Outside of work I make digital artwork and short videos, and I write about the engineering problems I find interesting.