Name: Krish Kumar
Role: Backend & Cloud-Native Engineer
Institution: Thapar Institute of Engineering and Technology (CSE Undergrad)
Core_competencies:
- Distributed Microservices & Async I/O
- Kubernetes Operators, Helm & GitOps
- High-Concurrency WebSockets & Pub/Sub
- Database Normalization (BCNF) & Atomic Transactions
Currently_building: CapstoneConnect β real-time student collaboration platform
Currently_learning: Service Meshes (Istio) & eBPF observability
Fun_fact: "Resolved race conditions across 100+ simulated concurrent requests β and lived to tell the tale"|
Languages Backend |
Cloud & DevOps Data & Observability |
|
Real-Time Student Team Discovery Platform Architected a 6-service microservices backend on Kubernetes with real-time bi-directional chat via Redis Pub/Sub and WebSockets β engineered for stability under 500+ concurrent connections.
|
|
βοΈ Nimbus β K8s Self-Service Deployment Engine Self-service PaaS letting developers ship apps without raw K8s YAML. Custom Kopf Operator in Python managing 3 CRDs, with Kaniko for rootless in-cluster image builds.
|
π Secure Share β Encrypted File Sharing Async, low-latency binary file distribution with a Rust backend and TypeScript client. Chunked transfer encoding and strict payload verification.
|
|
π₯ Hospital Management System β Healthcare Portal BCNF-normalized 8-table MySQL schema with 7 stored procedures, 3 functions, 6 triggers guaranteeing atomic appointment scheduling. Flask + SQLAlchemy service layer.
|
Predicts Remaining Useful Life of turbofan engines on NASA's C-MAPSS dataset using a Bidirectional LSTM with self-attention and Monte Carlo Dropout for uncertainty-aware predictions, deployed as an interactive Gradio dashboard.
|
|
π¦ Supply Chain Risk Intelligence β ML Classification Multi-class model predicting e-commerce delivery outcomes (late, on-time, advance, canceled) using SMOTE-balanced data and comparing XGBoost, Random Forest, KNN, and Logistic Regression with SHAP explainability, deployed via Gradio.
|
More projects on GitHub β |
- π‘οΈ Defensive & Idempotent Design β every state-changing endpoint tolerates partial failure and concurrency
- βΈοΈ Declarative Infrastructure & GitOps β manifests, Helm charts, and ArgoCD pipelines for repeatable deploys
- β‘ Database Efficiency First β normalized schemas, stored procedures for atomic ops, Redis caching
- π Telemetry & Observability β Prometheus + Grafana before incidents reach users



