Software Engineer Portfolio
Pradhyuman AroraSoftware Engineer
Software Engineer at Microsoft Defender for Cloud Apps, building high-throughput cloud infrastructure, distributed systems, and secure backend services at scale.
About Me
I'm a Software Engineer on the Shadow IT team at Microsoft Defender for Cloud Apps, focused on building reliable, high-throughput backend systems on Azure. I care about clean architecture, scale, and secure-by-default engineering.
What I Do
At Microsoft Defender for Cloud Apps, I own backend features for the Cloud Discovery engine that processes real-time logs via Azure Event Hubs to power shadow IT detection for 45k+ daily active users. I've led a zero-downtime migration of 500+ GB of production data from MongoDB to Azure Cosmos DB and built the Cloud App Catalog API from scratch.
More recently, I've been building an AI Copilot (MCP) for Shadow IT Discovery, integrating LLMs with high-throughput backend APIs and designing semantic caching with Redis Vector Search to cut inference overhead and latency.
How I Work
I believe good systems are simple, observable, and secure by default. I optimize for reliability and scale, invest in strong test coverage and automation, and enforce least-privilege access across the services I build. Every change should be measurable and safe to ship.
Core Values & Principles
Reliability at Scale
Systems should stay fast and correct under peak production load
Security by Design
Zero-trust and least-privilege access built in from the start
Continuous Learning
The stack keeps evolving, and so does the way I build
Professional Journey
Two years building production cloud infrastructure and security products at Microsoft, with earlier work spanning full-stack development and applied AI.
Software Engineer, Shadow IT
Key Contributions
- Owned backend features for the high-throughput Cloud Discovery engine, processing real-time logs via Azure Event Hubs to power shadow IT detection for 45k+ daily active users
- Led the zero-downtime migration of 500+ GB of production data from MongoDB to Azure Cosmos DB across 12 microservices, executing cutover in under 15 minutes with 0 data loss
- Built the Cloud App Catalog API from scratch as a core internal shared service, exposing risk and compliance metrics for 31k+ applications to multiple engineering teams across Microsoft
- Architected a Redis-based distributed cache (Cache-Aside pattern), slashing Azure Cosmos DB read overhead by 92% and optimizing downstream dependency latency under peak loads
- Engineered an AI Copilot (MCP) for Shadow IT Discovery, integrating an LLM layer with high-throughput backend APIs to enable natural language querying of SaaS security metrics
- Designed a semantic caching layer using Redis Vector Search, slashing LLM inference overhead by 40% and achieving sub-30ms latencies
- Drove zero-trust adoption across 60+ Azure Event Hubs by migrating to Managed Identities, and led SFI compliance execution hardening 200+ Azure resources with Network Security Perimeters
Impact
- Migrated 500+ GB of production data with 0 data loss and a <15 minute cutover
- Cut Azure Cosmos DB read overhead by 92% via a Cache-Aside Redis layer
- Reduced LLM inference overhead by 40% and token utilization by 85% for the Shadow IT Copilot
- Automated the Catalog update pipeline via Azure DevOps, cutting lifecycle cycles from 12 to 2 hours (83% reduction)
- Achieved 100% closure on 12 security KPIs and authored 200+ unit tests across 12 critical service paths
Tech Stack
Software Developer Intern
Key Contributions
- Built a React Native asset-digitization app integrating mobile camera and GPS, enforcing 500m geospatial validation between asset capture and form submission
- Integrated OwlViT-based object detection with a Flask backend to classify infrastructure assets into predefined categories and store detected assets with surveyor-provided metadata
Impact
- Enforced 500m geospatial validation for asset survey submissions
- Enabled infrastructure asset classification and storage with surveyor-provided metadata
Tech Stack
Summer SWE Intern
Key Contributions
- Built an AI-powered phishing payload generator for Microsoft Defender for Office 365
- Engineered a resilient dual-GPT pipeline via Augloop for fallback reliability
- Automated evaluation of generated payloads to enhance security threat detection
- Deployed production-ready code for Microsoft employees worldwide
Impact
- Delivered a dual-GPT phishing payload generator for Defender for Office 365
- Improved fallback reliability through a resilient dual-model pipeline
- Deployed code reaching 200K+ Microsoft employees
Tech Stack
Engage Mentee
Key Contributions
- Engineered ML-based music recommendation system with intelligent categorization
- Developed facial expression detection system using computer vision
- Built full-stack application with ReactJS frontend and Django backend
- Implemented real-time video analysis with machine learning algorithms
Impact
- Built music recommendation system with 5 category video suggestions
- Achieved 85% accuracy in facial expression detection
- Successfully integrated ML models with web application
Tech Stack
Skills & Technologies
An interactive map of the languages, cloud platforms, and tools I use to build and ship reliable backend systems at scale
Hover over skills to see connections • Click for detailed information
Featured Projects
Production systems I've designed and shipped at Microsoft Defender for Cloud Apps — spanning high-throughput APIs, data migration, distributed caching, and AI.
Cloud App Catalog API
Core Internal Shared Service
Built from scratch to expose risk and compliance metrics for 31k+ applications to engineering teams across Microsoft.
Shadow IT AI Copilot
LLM-Powered Security Assistant (MCP)
AI Copilot enabling natural language querying of SaaS security metrics, backed by high-throughput APIs and semantic caching.
MongoDB → Cosmos DB Migration
Zero-Downtime Data Migration
Led a zero-downtime migration of 500+ GB of production data across 12 microservices with 0 data loss.
Distributed Cache Layer
Cache-Aside Pattern with Redis
Architected a Redis-based distributed cache that slashed Cosmos DB read overhead by 92% under peak loads.
Impact at a Glance
Get in Touch
Interested in working together or have a question? Send me a message and I'll get back to you as soon as I can.