Projects

The projects I have developed from the past 2-3 year so far are included in this section, along with a concise overview of my work within these projects.

👉 zerch: Github

• A vector search engine for logs using Rust with cosine similarity, achieving 90% more accurate log retrieval compared to keyword-based search.
• Features:
  • ONNX-based embedding pipeline converting log lines into 384-dimensional vectors using all-MiniLM-L6-v2 model.
  • Processes 1,000+ logs/second with batch upserts to Qdrant.
  • MCP server with stdio/HTTP transports for AI agents like OpenClaw and Hermes.
  • JSON-RPC with embed text and search logs tools.
  • Integrated Groq API for AI-powered log summarization, reducing incident debugging time by 70%.
• Tech: Rust, ONNX Runtime, Qdrant, MCP, Groq API.

👉 dappergo: Github

• A lightweight distributed tracing library inspired by Google’s Dapper in Go for observing and debugging microservices.
• Features:
  • gRPC-based span collection system with buffered channel architecture.
  • Processes 10,000+ spans per second without blocking main threads.
  • Context-based trace propagation flowing across 15+ service boundaries with zero context loss.
  • 100% accurate dependency mapping.
  • Probabilistic sampling reducing overhead by 80% while capturing statistically representative traces with <5ms latency.
• Tech: Go, gRPC, protobuf.

👉 FlowCD: Github

• FlowCD is a lightweight GitOps controller for Kubernetes inspired by Argo CD.
• Features:
  • Quick sync with Git repo.
  • Automatically syncs manifests from your Git repository.
  • Deploy to local Kubernetes cluster and apply synced manifests to keep the cluster in sync with Git.
• Status: In active development — architecture and design decisions are ongoing.

👉 thanos-mcp-server: Site

• A custom Thanos MCP server that provides AI agents a single endpoint to query and analyze globally aggregated, long-term Prometheus metrics.
• Thanos does not have thier own mcp server so i built my own.
• Features:
  • PromQL query tool: execute PromQL queries against Thanos/Prometheus endpoints.
  • MCP integration: works with MCP-compatible clients like Cursor.
  • Global metrics access: query aggregated metrics across your monitoring infrastructure.
• Warning: NOT READY FOR PRODUCTION — experimental/development project.

👉 federated-k8s-demo: Github

• College project implementing federated learning across multiple K3s clusters using Flower and TensorFlow.
• Privacy-preserving: differential privacy (TensorFlow Privacy), homomorphic encryption (TenSEAL), HTTPS/TLS, and OAuth2/JWT.
• Uses FedAdam/FedProx strategies, model compression (mixed precision, TF Lite quantization), and Kubernetes Jobs for on-demand training.
• Monitoring with Prometheus + Grafana. Focus: chest X‑ray classification for privacy-sensitive medical training.

👉 SkinSprite: Github

• Built an AI-driven Personal Skin Health Assistant for diagnosing skin diseases. Users can upload images and receive responses along with details of nearby dermatologists.
• Used the VGG19 Convolutional Neural Network architecture for accurate disease identification.
• Developed the app with Java and Spring Boot, and the website with React.js, Node.js, and MongoDB. Used ngrok for tunneling to the ML model API.

👉 FlavorFleet: Github

• Developed FlavorFleet, a food order site with Node.js and MongoDB.
• Containerized the application with Docker and pushed it to Docker Hub.
• Deployed the site on Kubernetes for seamless deployment.

👉 CI-CD-Pipeline: Github

• Created a simple Golang application with a basic web server.
• Dockerized the application and tested it locally.
• Set up a CI/CD pipeline using Jenkins, including Docker for building and managing Docker images.

👉 microservices-backend: Github

• Developed three microservices (User Service, Restaurant Service, and Delivery Agent Service) for a food delivery app using Node.js and PostgreSQL.
• Containerized each microservice with Docker and set up Docker Compose for streamlined multi-container management.