Mittapally Harshith
Two years of professional experience spanning .NET backend development and cloud/DevOps engineering — containerised deployments on Kubernetes, CI/CD pipelines, and infrastructure as code with Terraform. I'm now deepening my AWS specialisation toward the Solutions Architect – Associate certification, and keep shipping real systems along the way: an explainable ML platform deployed on EC2/S3 for my MSc dissertation, and a production exam-prep app I designed, built, and operate myself.
Professional experience
Two years across backend development and cloud/DevOps engineering before starting the MSc.
Junior DevOps Engineer
- Deployed and maintained containerised .NET applications on Amazon EKS using Kubernetes manifests and Helm charts
- Built and maintained Docker images for .NET applications, and Jenkins CI/CD pipelines to automate build and deployment workflows
- Implemented rolling and blue/green Kubernetes deployment strategies, and troubleshot pod failures, networking issues, and pipeline errors across the stack
- Supported enterprise application modernisation through containerisation and migration toward microservices
Software Engineer
- Developed backend applications and REST APIs using C#, ASP.NET Core, MVC, and Web API
- Built and maintained RESTful services for enterprise applications, and optimised SQL queries and database schemas for performance
- Investigated and resolved production defects through debugging and root-cause analysis within Agile development teams
I'm a DevOps and Cloud Engineer with two years of professional experience — a year building backend services in C# and ASP.NET Core at Cognizant, followed by a year deploying and operating containerised .NET applications on Kubernetes at Genex Corporate Services, where I worked daily with Docker, Helm, Jenkins, and Terraform. I moved to Dublin to do an MSc in Computer Science and formalise the cloud side of that experience, with a deliberate focus on AWS.
My MSc dissertation combined explainable ML with real AWS deployment: a network intrusion detection model trained to near-perfect accuracy, explained with SHAP so a security analyst can trust its decisions, then shipped behind a FastAPI service and Streamlit dashboard on EC2, backed by S3 for storage. I'm currently studying for the AWS Solutions Architect Associate exam and building aws-knowledge-app — a full exam-prep platform I designed, built, and operate myself — as hands-on practice for exactly the kind of deploy-and-maintain work I want to do professionally.
Academic background
MSc Computer Science
Coursework spanning cloud programming & architecture, parallel & distributed programming, generative AI / deep learning, data mining & machine learning, mobile development, and advanced networks & cybersecurity. Dissertation: Intelligent Network Intrusion Detection Using Explainable Machine Learning and Cloud Deployment — see below.
BTech, Computer Science Engineering (Data Science)
Full-stack foundations (HTML/CSS/JS, Angular, Node.js, Express.js) alongside Python, Java, C, R, and database management (SQL, MySQL). Final-year projects in sentiment analysis, smart parking systems, and ML-based crop-yield prediction.
MSc Dissertation
An end-to-end intrusion detection system: trained, explained, and deployed on real cloud infrastructure — not just a notebook.
Intelligent Network Intrusion Detection Using Explainable Machine Learning and Cloud Deployment
Built a binary network intrusion classifier on the CICDDoS2019 flow dataset, comparing Logistic Regression, Decision Tree, Random Forest, and XGBoost. Random Forest won on every metric. Used SHAP to explain both global feature importance and individual predictions, so a security analyst can see why traffic was flagged, not just that it was. Deployed the final model behind a FastAPI inference API and a Streamlit dashboard on AWS EC2, with datasets and prediction outputs streamed to and from S3 — deliberately designed so multi-gigabyte CSVs never have to sit on the EC2 instance's own disk.
What it does
- Classifies network flows as benign or attack traffic in real time
- Explains every prediction with SHAP — global and per-record reasoning
- Serves predictions via REST API (/predict, /predict-csv, /model-info)
- Streams large CSVs through S3 instead of holding them in EC2 memory
- Writes prediction outputs back to S3 for downstream analysis
Stack
aws-knowledge-app
A full exam-prep platform for the AWS Solutions Architect Associate — designed, built, and operated by me, end to end.
AWS SAA exam-prep platform
Covers every service in the exam syllabus with structured concepts, scenario walkthroughs, AI-graded free-text answers, a flashcard-style quiz engine with spaced review, full exam simulation mode with a 150-question capstone format, and a tiered rank system that tracks mastery per service. I own the whole lifecycle — architecture, deployment, and every update — on a live EC2 instance behind my own domain.
Selected projects
DevOps and AWS infrastructure work, plus coursework across full-stack, ML, mobile, and systems programming.
Cloud-Native Deployment Platform
DevOps · KubernetesProvisioned cloud infrastructure with Terraform and deployed containerised .NET services onto Kubernetes, with GitHub Actions driving the build-and-deploy pipeline and Prometheus/Grafana wired up for container metrics and troubleshooting.
AWS High-Availability Web App
AWS · InfrastructureDeployed Apache web servers on two EC2 instances split across Availability Zones behind an Application Load Balancer, with Route 53 for DNS routing and CloudWatch for monitoring. Automated instance stop/start scheduling with EventBridge and Lambda to control cost.
TweetSpace
Cloud ArchitectureA Twitter-style social platform: FastAPI backend, MongoDB for users/tweets, Firebase Authentication, and Azure Blob Storage for media. Soft-deletes, retweet graphs that preserve the original author, and a follow/follower system modelled directly in each user document.
Oracle v1.4.0
Team · AgileAn AI-powered RAG document assistant built across sprint cycles with a live client stakeholder: PDF ingestion, semantic retrieval, vector storage, citation-backed chat responses, voice interaction, and an admin dashboard for corpus management.
AG News Text Classification
Deep LearningBuilt a BiLSTM-with-Attention classifier from scratch (84.82% test accuracy), then fine-tuned DistilBERT on the same four-class news task (92.33% accuracy). Interpreted both models with token occlusion and t-SNE embedding visualisation.
Mobile Price Classification
Data Mining & MLFull pipeline on a mobile-pricing dataset: EDA, feature engineering, four baseline models tuned with RandomizedSearchCV, KMeans clustering, and Apriori association-rule mining to surface pricing patterns tied to hardware specs.
Expenses Tracker
Android · SoloA Kotlin + Jetpack Compose budgeting app with zero XML layouts and zero external libraries by design. State-lifting architecture, hand-built bar and pie-chart analytics, and a clean git history from initial setup through bug fixes.
Parallel Palindrome Finder
Systems · HPCA C program finding palindromic substrings using both MPI and OpenMP, benchmarked across 1–8 processes to measure real throughput scaling — direct hands-on experience with distributed and shared-memory parallelism.
Technical toolkit
Cloud & Infrastructure
Containers & CI/CD
Monitoring & Observability
Backend & Databases
ML & Data
Languages
AWS Certified Solutions Architect – Associate
Let's talk.
Open to Cloud, DevOps, and Platform Engineering roles — and any role with a path to long-term sponsorship.