Vishal Lakshmi Narayanan

Developed and published an open-source Python package for tail risk modeling with specialized loss functions (LaR, CVaR) and hybrid meta-learning ensembles. Improves tail coverage ratio by 76% over baseline models, available on PyPI with comprehensive documentation and scikit-learn compatibility.

Built an AI-powered resume-job matching platform using semantic embeddings and transparency scoring to match candidates based on capability rather than keywords. Developed dual portals for job seekers and recruiters, with automated candidate ranking, multi-criteria scoring, and AI-generated job-specific resumes via LaTeX.

Built a risk-aware ensemble framework for catastrophic insurance loss prediction using Loss-at-Risk (LaR) and CVaR optimization. Developed tail-focused meta-learners and a hybrid model that improved extreme-tail MSE by 3.2% and CVaR residuals by 10.25% over baseline models on 69K+ insurance claims.

Celestia AI turns maps into living worlds. Drop a marker anywhere and watch a holographic mascot bring the place to life with cinematic video loops, six fact cards, and real-time narration. Instead of reading static text, users step into a guided storytelling experience where culture, heritage, and history feel tangible.

An AI-powered learning assistant that transforms dense concepts into short, dialogue-based explainer videos with quizzes, streaks, and badges. If a student doesn’t understand a concept the first time, SlideSage instantly regenerates with fresh analogies or examples until it clicks — making learning adaptive, fun, and addictive.

Engineered an AI-driven platform that generates optimized multi-day travel itineraries with real-time route refinement. Integrated cost estimation, weather forecasts, and interactive map visualizations, improving planning accuracy by 30%. Won People's choice Award for this project at Hackazona v0.1.

Built an AI job-matching portal with semantic similarity scoring and automated resume parsing for 100+ applications. Developed secure recruiter/applicant dashboards with authentication, job posting, and real-time resume uploads.

Developed an n8n AI agent that automates email and scheduling via Gmail, Calendar, and Sheets. Implemented contact verification workflows ensuring email addresses are validated before sending messages.

PhishGuard is a Chrome extension that uses AI and VirusTotal to detect phishing websites in real-time. Click the PhishGuard icon in the Chrome toolbar, choose “Analyze This Page,” and view the risk level and detailed recommendation in the popup.

Built a Flask-based web application using EfficientNetB0 for medical image classification. Designed for low-cost, smartphone-compatible retina scans to support early diagnosis in rural healthcare settings.

Extracted network traffic features and trained machine learning models for real-time cybersecurity applications. Built a threat classification pipeline using supervised learning techniques.