const suhani = {
  role:     "AI/ML Engineer · MERN Stack Developer",
  edu:      "Dual Degree ECE, NIT Hamirpur (2028)",
  focus:    ["LLMs", "RAG Pipelines", "Scalable Backend APIs"],
  building: "intelligent systems + the infrastructure behind them",
  learning: ["LangGraph", "Advanced RAG", "Go", "System Design"],
  motto:    "ship fast, build things that actually work",
};

I build at the intersection of Artificial Intelligence and Backend Engineering — designing LLM-powered systems and the scalable APIs that serve them. I like turning rough ideas into production-ready products.

A production-ready MERN platform with dual interfaces for customers and fleet owners.

• 🔐 JWT auth + role-based access control (user / owner)
• ⚡ Race-condition-safe booking via date-overlap logic — 0% booking conflicts under load
• 📈 Optimized REST APIs handling 100+ concurrent requests
• 🤖 Google Gemini integration for AI-generated car descriptions — 80% faster listing creation
• 🖼️ ImageKit CDN + WebP compression — 70% smaller payloads, 2x faster load

React · Node.js · Express · MongoDB · Google Gemini · ImageKit

A full-stack disaster management platform developed during my Software Engineering Internship at IIT Roorkee, featuring real-time environmental monitoring, analytics, and secure multi-tenant access.

• 🔐 Designed a multi-tenant Role-Based Access Control (RBAC) system for isolated organizational workspaces
• 📊 Built real-time monitoring dashboards with live sensor analytics and risk visualization
• 📈 Developed a cross-correlation analysis module for statistical comparison of environmental sensor data
• 📉 Created interactive historical charts with CSV export for streamlined research workflows
• ⚡ Optimized backend performance through connection pooling, chunked data loading, and modular Django architecture

Django · Python · JavaScript · HTML/CSS · Chart.js · SQLite

A multilingual AI pipeline preserving cultural context in machine translation.

• 📚 Processed 500+ Hindi idioms through a multi-model AI workflow
• 🔍 LaBSE embeddings + cosine similarity to filter semantic noise
• 🧩 Scikit-learn ensemble aggregation to combine LLM outputs and preserve meaning

Python · Groq API · Llama Models · LaBSE · Scikit-Learn · RAG