Building production-grade AI systems, cloud infrastructure, and secure software — end to end.
┌─────────────────────────────────────────────────────────────────────────────┐
│ Dharshan Kumar J │
│ │
│ I don't just build applications. I design systems. │
│ AI agents, secure backends, cloud pipelines, ML models — │
│ engineered for production, not just demos. │
│ │
│ Domain crossover: Software × AI/ML × Cybersecurity × DevOps │
└─────────────────────────────────────────────────────────────────────────────┘
↳ Click to expand project portfolio
End-to-end DevOps lifecycle: Terraform provisioning → Docker containerization → Jenkins CI/CD → Azure production deployment. Built to demonstrate infrastructure-as-code discipline from zero to running.
Production-grade ML classification system: ensemble pipeline achieving 98.2% accuracy, FastAPI REST layer, PostgreSQL persistence, Prometheus + Grafana observability, and full CI/CD. Not a notebook — a system.
End-to-end Retrieval-Augmented Generation: PDF ingestion → Sentence Transformer embeddings → ChromaDB → BM25 + semantic hybrid retrieval → Cross Encoder reranking → 3-stage confidence validation → Llama 3.2 via Ollama. Zero cloud APIs. Runs offline.
AI-powered resume comprehension — not parsing, understanding. Leverages Mistral 7B via OpenRouter to analyze candidate depth across skills, projects, and expertise. Built for recruiters who want signal, not noise.
Research-grade dataset from verified DeFi exploits + safe production contracts. Slither static analysis → feature extraction → Gradient Boosting → SHAP explainability. 98.49% accuracy. Paper accepted at IEEE — publication in progress.
↳ Skills by domain
Frontend
Backend
Database
Tools
DevOps
App Dev
Cloud
AI / ML
The best systems aren't just functional — they're intentional.