Hello, I'm Kasi Viswanath

Problem: Late detection of Gestational Diabetes in low-resource clinics.

Solution: Built a lightweight ML screening model using clinical features for early-risk prediction.

TECH: Python, Scikit-learn, Pandas, ML Classification

Problem: Hidden lifestyle indicators of stroke often go unnoticed.

Solution: Designed a high-recall predictive model prioritizing early risk detection.

TECH: Python, Scikit-learn, Data Preprocessing, ML Models

Problem: Classical ML struggles with high-dimensional oncology data.

Solution: Compared Deep Learning with hybrid Quantum ML models to test feature separability.

TECH: Python, PyTorch, Qiskit, Neural Networks

Problem: Teachers lack real-time insight into classroom engagement.

Solution: Built a privacy-aware analytics pipeline converting behavior signals into engagement metrics.

TECH: Python, OpenCV, Data Analytics, Visualization, Fast API

Problem: Manual handling of event registrations and coordination for a tech fest.

Solution: Developed and deployed a centralized web platform for registrations and participant flow.

TECH: HTML, CSS, JavaScript, Vercel, Responsive UI

Problem: Manual resume shortlisting is slow and inconsistent.

Solution: Created an NLP pipeline using TF-IDF and cosine similarity for semantic ranking.

TECH: Python, NLP, Scikit-learn, TF-IDF, Streamlit

Problem: Gaming requires external controllers or hardware.

Solution: Built gesture-controlled games using real-time hand tracking via webcam.

TECH: Python, OpenCV, MediaPipe, Computer Vision

Problem: Manual timetable creation causes frequent scheduling conflicts.

Solution: Designed a constraint-based scheduling system handling rooms, faculty, and slots.

TECH: Python, Logic Algorithms

Problem: Satellite weather images show current clouds, but predicting future cloud movement precisely is complex and compute heavy.

Solution: Designed a spatio-temporal diffusion model pipeline to predict next cloud frames using multi-sequence satellite imagery and CNN inference accelerated on NPU.

TECH: Diffusion Models, CNN, Satellite Imagery, Python, NPU Acceleration, Computer Vision

Problem: Mental health risks are usually detected only after visible symptoms, missing early intervention opportunities.

Solution: Built a pattern-analysis AI that evaluates responses to a dynamic 10-question behavioral set and predicts early mental stress indicators using ML classification.

TECH: NLP, ML Classification, Pattern Analysis, Python, Behavioral Data Modeling

Problem: Traditional models operate as black boxes, lacking clear physiological reasoning.

Solution: Built a dual-inference pipeline combining classical ML with quantum-inspired modeling for transparency.

TECH: Python, Scikit-learn, PennyLane, SHAP, FastAPI

Problem: High-accuracy models fail real-time constraints on edge devices.

Solution: Controlled pipeline measuring input resolution against latency, boosting FPS with minor accuracy loss.

TECH: YOLO, OpenCV, Python, Edge Benchmarking

Problem: Insufficient centralized visibility for hackathon progress and submissions.

Solution: Created a role-aware AWS-native monitoring platform for centralized tracking.

TECH: AWS Amplify, Lambda, API Gateway, DynamoDB