Portfolio

All Projects

Architectures, platforms, and solutions across Azure, AI, DevOps, Security, and FinOps.

Projects

Project Gallery

A selection of architectures, platforms, and solutions delivered across industries.

Marina Meals — Bilingual Voice Ordering on Azure Voice Live

An end-to-end voice ordering experience for a fictional Chennai restaurant, built on the Azure Voice Live API in Azure AI Foundry. Customers speak naturally in English or Tamil to browse the menu, check live prices, find a branch, and place a pickup order — no forms, no ordering flow to navigate. The design point is grounding: restaurant facts are never generated from model memory. Prices, availability, store hours and order totals all come from backend tools that re-read the current restaurant data on every call, and the server — not the model — calculates money.

Real-time bilingual conversation — PCM16 mono at 24 kHz with semantic voice activity detection and barge-in, English-first with Tamil replies reserved for complete Tamil utterances

Tool-grounded answers — menu, price, availability, branch and order facts resolve through backend functions that re-read live JSON on every call, so a price edit is reflected mid-session without a restart

Server-side order totals — the model never calculates money or invents confirmation numbers

+4 more

Azure Voice Live APIAzure AI FoundryMicrosoft Entra IDAzure App Service+11

AIAzure

Frontier Boardroom — A Multi-Model AI C-Suite

Five AI agents act as a virtual C-suite — CEO, CFO, CMO, CTO and Legal — each running on a different frontier model and debating a real business decision in a live Teams-style conversation. OpenAI, xAI, Meta and Mistral route through Microsoft Foundry; Anthropic through Azure Databricks Mosaic AI. Any seat can be reassigned or hot-swapped mid-session, and every persona answers from one shared Foundry IQ knowledge base with inspectable citations.

Five frontier models in one live debate — three seats on Microsoft Foundry, two on Azure Databricks Mosaic AI

Any seat reassigned or hot-swapped mid-session, with no change to persona or orchestration code

Grounded, not improvised — shared Foundry IQ retrieval with citations users can open and inspect

+5 more

Python 3.12FastAPIPydantic v2WebSockets+18

AzureAI

Portfolio Website — Azure Cloud Architecture

Designed and built a production-grade portfolio website on Azure, treating it as a reference architecture. Runs on Azure App Service (B1, Central India) with a RAG-powered AI chatbot via Azure AI Foundry, GitHub Actions CI/CD, passwordless Managed Identity auth, and Application Insights monitoring. Every infrastructure decision is documented at /architecture.

Full Azure cloud infrastructure — App Service, AI Foundry, Blob Storage, App Insights, Managed Identity

RAG chatbot grounded on portfolio content using gpt-4o + file_search vector store

Zero API keys in production — Managed Identity with Cognitive Services User + Azure AI Developer RBAC

+1 more

Next.js 16Azure App ServiceAzure AI FoundryAzure Blob Storage+4

Azure

Cloud-Ready Academy — Azure Skilling at Scale

Led end-to-end design of a structured Azure skilling initiative. Built hands-on labs, capstone projects, and learning modules to upskill diverse engineering audiences on cloud-native development and architecture.

Achieved 95% skills enhancement rate among participants

Scaled cloud learning to thousands of engineers

Built reusable lab content now used across multiple cohorts

Microsoft AzureAzure LabsGitHubHands-on Learning

SupportIQ — Generative AI for Support Automation

Built a Generative AI solution on Azure to automate support workflows for engineering teams. Leveraged Azure OpenAI and AI Foundry to intelligently triage, route, and resolve common support issues without manual intervention.

Reduced manual effort by ~1 hour per engineer per day

Improved support engineer efficiency by 14%

Delivered measurable operational and customer experience improvements

Azure OpenAIAzure AI FoundryPythonAzure Functions+1

Azure

Azure Landing Zone for Startups

Designed a right-sized Azure Landing Zone for a high-growth startup — hub-spoke networking, Azure Policy governance, identity baseline, and a self-service developer platform. Balanced enterprise-grade security with startup velocity.

New environment provisioning: 2 weeks → 2 hours

Security baseline achieved from day one

Passed investor due diligence cloud security review

AzureTerraformAzure PolicyEntra ID+1

Agentic AI Workflow Automation

Designed and implemented an Agentic AI system using Azure AI Foundry and Semantic Kernel that autonomously handles multi-step business workflows — from data ingestion and analysis to report generation and stakeholder notification.

Automated 12 previously manual end-to-end workflows

Reduced processing time by 80% on targeted workflows

Zero human intervention required for standard workflow execution

Azure AI FoundrySemantic KernelAzure OpenAIPython+1

DevOps

DevOps Platform Engineering

Built an internal developer platform enabling self-service infrastructure provisioning, CI/CD pipeline templates, and standardized observability across multiple engineering teams.

Developer onboarding time reduced from 3 days to 2 hours

CI/CD pipeline adoption increased to 100% of teams

Deployment lead time reduced by 65%

Azure DevOpsTerraformGitHub ActionsAzure Monitor+1

Azure

Cloud-Native Platform Modernization

Led the modernization of a legacy monolithic application to a cloud-native, containerized microservices architecture on Azure Kubernetes Service. Implemented GitOps, observability, and zero-downtime deployment strategies.

Deployment frequency increased from weekly to daily

Mean time to recovery reduced from hours to minutes

Infrastructure costs reduced by 40%

AKSDockerTerraformGitHub Actions+2