🧠 Projects
Exploring secure, privacy-conscious, and trustworthy AI systems — from conversational agents to secure inference pipelines.
🤖 AI HealthBot — Conversational Medical Assistant
Overview:
AI HealthBot is a conversational AI system designed to help users describe symptoms, assess potential conditions, and receive precautionary guidance. Built with LLMs, NLP, and multi-agent coordination, it demonstrates secure deployment and privacy-conscious interaction design.
Core Features:
- 🩺 Symptom analysis and probable disease matching
- ⚕️ Precaution and risk-level assessment
- 💬 FAQ-style health information responses
- 🧩 Vector-based retrieval using FAISS and LangChain
- 🧠 Contextual multi-turn dialogue with memory retention
Tech Stack: Python, Transformers, LangChain, FAISS, Gradio, Hugging Face Spaces
Security & Privacy Focus: Incorporates differential privacy principles and input sanitization to prevent data leakage and ensure ethical AI interaction.
⚙️ Secure Model Inference Pipeline — From Model to Production, Safely
Overview:
A full-lifecycle MLOps pipeline that integrates security validation, adversarial testing, and automated model promotion. The system ensures only verified and robust models progress through CI/CD using Jenkins and MLflow.
Pipeline Workflow:
- 🧰 Model Scanning: AV and modelscan validation for embedded threats
- 🧪 Adversarial Testing: Stress-tested with Adversarial Robustness Toolkit (ART)
- 🧾 Integrity Verification: SHA-256 hashing and MLflow registration
- 🚦 Secure Promotion: Auto-promotes passing models to Staging
Tech Stack: Python, Jenkins, MLflow, Docker, Bash, ClamAV, ART, Flask, AWS S3
Security Highlights:
✅ Model provenance tracking via MLflow
✅ Containerized isolation for test environments
✅ Secure secrets handling in CI/CD
✅ Integration-ready with Secure Model Context Protocol (SMCP)
Impact:
This pipeline enforces defensive MLOps — embedding security gates, provenance, and compliance directly into the ML lifecycle.