Ayush Narayanaswamy
Management Engineering · UWaterloo
Product · Data · Systems
Management Engineering · UWaterloo
Product · Data · Systems

LocalStudent
UWaterloo
UW Formula Electric
MECH
— Medical Envoys for Community Health
Contactless vitals monitoring and triage platform using remote photoplethysmography (rPPG) to extract heart rate, HRV (SDNN, RMSSD), and respiratory rate from smartphone camera input. Streams frames from a Flutter app to a FastAPI backend where MediaPipe FaceMesh + CHROM isolate blood volume pulse signals, with FFT-based analysis for real-time physiological estimation. Integrates a multi-agent Claude pipeline to combine vitals with adaptive symptom intake and generate structured urgency classifications with plain-language summaries.
Biometric-adaptive study platform capturing heart rate, HRV, breathing rate, and blink signals from smartphone camera via Presage SmartSpectra SDK. Streams physiological data over WebSockets to a locally-hosted Qwen 2.5 LLM (Ollama) which classifies cognitive states — flow, fatigue, frustration, disengagement — and generates adaptive study recommendations in real-time. Improved measured study efficiency by 75%.
Multi-persona conversational AI platform engineered on the Google Gemini API, delivering adaptive emotional support through three distinct relational behavioral profiles: therapist, companion, and ex-partner. Implements dynamic prompt engineering with contextual session memory to maintain persona coherence across extended conversations. Features real-time streaming response architecture, emotional state inference from user input patterns, and a Next.js App Router frontend with custom conversation state management hooks.
Chrome extension performing real-time credibility analysis on long-form media sources during active browsing. Implements live DOM parsing and metadata extraction combined with bias-detection heuristics and citation-density scoring. Lightweight NLP models evaluate credibility signals across 1,000+ analyzed pages, surfacing inline reliability indicators and source bias flags — with natural language explanations via a locally-hosted Ollama inference layer.
Comprehensive personal finance management system engineered entirely in Excel VBA. Supports structured data management for 5,000+ financial records with automated categorization pipelines and validated data-entry UserForms enforcing strict schema integrity. Features 10+ automated analytical dashboards — spending trend analysis, net cash flow visualization, investment projections with trained exponential growth models, and budgeting scenarios.
University of Waterloo's largest AI club — working on projects at the intersection of machine learning, applied AI research, and engineering. Engaged in workshops, research reading groups, and collaborative ML builds.
Collaborative learning community focused on data analysis, statistical modelling, and applied data pipelines. Participating in case competitions, workshops, and project teams applying data science to real-world problems.
Elected by 100+ Management Engineering peers to serve as the department's Engineering Society representative — attending faculty meetings, surfacing student concerns, and coordinating academic feedback between students and administration.