Saratchandra Patnaik
Architecting Intelligent Systems at Scale
I bridge the gap between heavy backend infrastructure and modern AI. With production experience at Amagi Media Labs and a Master's from Arizona State University, I specialize in deploying RAG systems, Cloud-Native microservices, and secure cryptographic protocols.
Ex-Amagi | ASU '25 | AWS & Python Expert
Professional Experience
Software Engineer (Cloud & DevOps)
Amagi Media Labs | Bengaluru, India
Key Stack: AWS EKS, FastAPI, ArgoCD, Terraform, GenAI
Engineered high-throughput distributed systems for real-time media streaming and architected intelligent fault-tolerance workflows that eliminated 60% of operational overhead.
Scaling Cloud-Native Infrastructure (AWS & Kubernetes)
The Infrastructure: Owned the lifecycle of a high-availability microservices platform hosting 15+ critical services on AWS EKS.
Reliability Engineering: Transitioned deployment to GitOps using ArgoCD, ensuring version-controlled and reproducible infra. Achieved 99.9% release stability and increased deployment frequency by 30%.
Incident Response (DRI): Resolved a critical streaming outage by tracing concurrent inbound streams across distributed nodes, restoring full service availability within minutes.
High-Performance Backend & Media Streaming
Protocol Optimization: Engineered real-time streaming workflows for WebRTC and WebRTS clients. Optimized H.264/H.265 codecs and AWS MediaLive configurations to maximize device compatibility, reducing playback errors by 95%.
Backend Design: Built asynchronous, non-blocking REST APIs using FastAPI (Python) to handle high-concurrency requests, reducing data processing latency.
Hardware Integration: Integrated StreamDeck client systems to let broadcast operators control media dashboards in real time.
DevOps Automation & Infrastructure as Code (IaC)
Eliminating Toil: Automated customer onboarding with Terraform and Python to provision VPCs, subnets, and customer environments instantly.
Impact: Cut manual setup efforts by 90%, turning a multi-day process into a one-click operation.
Quality Assurance: Automated CDN configuration checks to proactively catch delivery issues, reducing CDN failures by 95%.
Architecting AI-Driven Observability & ML Pipelines
The Problem: Root cause analysis for server failures was manual and slow, often involving hours of sifting through raw logs.
The Solution: Engineered a custom GenAI Observability Pipeline that ingested terabytes of server logs into Large Language Models (LLMs). The system semantically analyzed error patterns to predict failures before downtime.
Impact: Automated root cause analysis reduced manual debugging time by 60%, allowing the team to focus on feature development.
Capsequo Project: Led deployment of Capsequo, a Python-based ML captioning pipeline. Optimized real-time API integration to reduce video stream processing latency by 36%, improving live broadcast viewer experience.
Software Engineer Intern
Blueplanet Solutions Inc. | India
Tech Stack: MySQL, PHP, JavaScript, Linux System Logs
Optimized full-stack performance for the "Campus Club" portal, serving 1,000+ users. Reduced search latency by 60% and resolved critical memory leaks.
Database Performance Optimization
The Context: As the "Campus Club" portal grew to 1,000+ profiles, search queries degraded to 3+ seconds, hurting UX.
The Action: Analyzed MySQL execution plans, refactored monolithic queries into optimized stored procedures, removed redundant joins, and implemented proper indexing.
The Result: Reduced server-side query execution time by 50%, enabling sub-second data retrieval.
Frontend Modernization & UX
The Action: Built a responsive, asynchronous search UI using JavaScript (AJAX) and PHP, replacing full-page reloads.
The Result: Improved time-to-result by 60% and reduced server bandwidth usage.
Stability & Root Cause Analysis
The Challenge: Intermittent crashes and out-of-memory errors during peak usage.
The Debugging: Parsed web server logs to trace a memory leak caused by unclosed database connections in legacy PHP.
The Result: Patched connection handling to ensure cleanup, improving stability by 30% and eliminating recurring crashes.
Graduate Teaching Assistant / Grader
Arizona State University | Tempe, AZ
- ▹ Evaluated graduate-level implementations of Applied Cryptography protocols, focusing on security standards and encryption logic.
- ▹ Mentored students on securing distributed systems against vulnerabilities.
AI Solutions Architect (Independent)
Passion Projects
- ▹RAG System: Architected a 'My Personal Agent' using ChromaDB and GPT-4, utilizing Clean Architecture to decouple LLM providers.
- ▹Multimodal AI: Integrated GPT-4 Vision for real-time visual diagnostics and built an ATS Resume Optimizer using vector embeddings.
- ▹High-Performance Computing: Engineered a multi-threaded UDP server in C++ with thread pooling and mutex locking to simulate real-time telematics transmission.
Technical Skills
AI & Machine Learning
LLMs (GPT-4, Claude, OpenAI API), RAG (Retrieval-Augmented Generation), LangChain, Computer Vision, PyTorch, TensorFlow, Scikit-learn, Pandas, NumPy, Matplotlib
GPU & High-Performance Computing
CUDA Programming, PyCUDA, Shared Memory Optimization, Loop Unrolling, Tiling, Grid-Block-Thread Architecture
Cloud & DevOps
AWS (EKS, Lambda, S3, EC2, Greengrass, Load Balancer), Kubernetes, Docker, GitOps, ArgoCD, Jenkins, GitHub Actions, Linux, Terraform, Azure (Fundamental)
Languages
Python, C++, Java, Go, TypeScript, JavaScript, SQL, PHP, R, Kotlin, C
Backend & Web Frameworks
FastAPI, Node.js, React.js, Express.js, Flask, REST APIs, Microservices, HTML/CSS, Bootstrap, Android Studio
Databases
PostgreSQL, Redis, ChromaDB (Vector DB), MySQL, Google Firebase, MariaDB
Developer Tools
Cursor (AI Editor), GitHub Copilot, Jira, Grafana, Shell Scripting
Projects
My Personal AI Agent (RAG)
A sophisticated RAG (Retrieval-Augmented Generation) application serving as a personal AI assistant. Features vehicle diagnostics from images, resume analysis & ATS optimization, and intelligent resume building. Built with Python, ChromaDB, OpenAI/Anthropic APIs, and vision AI integration.
View on GitHubEdge AI Face Recognition
Stack: AWS IoT Greengrass, Lambda, MQTT, PyTorch
Built a real-time "Smart Camera" system deploying MTCNN and FaceNet models directly to the edge. Engineered a custom deployment pipeline to run raw PyTorch models in a pip-free AWS Lambda environment. Orchestrated asynchronous communication between Edge and Cloud using MQTT and SQS.
View ArchitectureGPU Accelerated Image Processing
Stack: CUDA, PyCUDA, Python, C++
Achieved 20x speedup over CPU implementations by engineering a parallel 2D Gaussian Filter. Optimized kernel performance using Shared Memory Tiling, Loop Unrolling, and Grid-Stride Loops to minimize global memory latency. Validated image fidelity using PSNR and SSIM metrics.
View on Google ColabUDP Multi-Threading System
High-performance telecom component simulation that receives encrypted custom packets via UDP, processes them using multi-threading (producer-consumer pattern), and reports statistics via POSIX shared memory (IPC). Demonstrates advanced C++ system programming, network protocols, and concurrent processing.
View on GitHub