I’m a Computer Science graduate from Chandigarh University (2024) with IEEE-published research in multimodal AI. My paper — AI Narratives: Bridging Visual Content and Linguistic Expression — combines an InceptionV3 CNN encoder with a Transformer-based decoder for image captioning, trained on COCO.
Before Node2.io, I interned as an ML Engineer at Quicksilver Technologiesin 2023, where I built Python-based automation scripts and reusable data processing pipelines for ML evaluation, validation, and backtesting workflows.
Today I’m the AI Engineer at Node2.io, a Canada-based AI-native cloud platform for intelligent property and infrastructure management. Day-to-day I ship FastAPI services, PostgreSQL data workflows, and LLM-based automation for operational reporting and documentation — alongside CI/CD pipelines, containerized deployments, and Linux-side reliability work that keeps the multi-tenant production environment honest.
Outside Node2.io I build the same way in public. Fraud Radar is a tier-1-style fraud-detection platform (FastAPI + XGBoost + SHAP, PR-AUC 0.9327, p50 3.7 ms). Unhosted is a Rust runtime that pools heterogeneous hardware into a single LLM inference cluster. Three live ML applications round it out — diabetes, heart-disease, and 38-class plant-disease classifiers serving real users.
→
Production over benchmarks
Models that hold up under real load — recall-tuned for what the use case actually costs, not leaderboard accuracy.
→
Full-stack ML ownership
Data pipelines, model architecture, training, evaluation, inference APIs, and the UI in front of them — owned end to end.
→
Research plus engineering
IEEE-published in multimodal AI. The same depth applied to applied engineering work at Node2.io.
