karthik Reddy

KneeVision-X — Research-Grade Knee MRI A

Python, PyTorch, FastAPI, React, DICOM, Multimodal AI 2026 – Present – Built a multi-plane, multi-pathology knee MRI AI system with Grad-CAM explainability, uncertainty estimation, calibration, and automated clinical-style reporting. – Designed Pathology-Aware Report Contrastive Learning (PARCL) for MRI–radiology-report alignment and label- efficient multimodal representation learning. – Integrated MRNet, SKM-TEA, fastMRI, RSNA 2026, and independent OA datasets with leakage checks, robustness testing, and reproducible evaluation. – Achieved 0.832 macro AUROC on knee OA severity grading and developed a full-stack FastAPI + React research dashboard.

AgriIntel AI — Intelligent Crop Yield Prediction & Smart Farming Platform

Developed an AI-powered agricultural decision-support platform integrating crop recommendation, yield prediction, plant disease detection, soil/weather intelligence, fertilizer guidance, market intelligence, and conversational AI. – Built a HistGradientBoosting yield prediction model achieving R2 0.846, MAE 331.2 kg/ha, and RMSE 655.7 kg/ha on an unseen 2010–2017 test set. – Developed a ResNet-18 plant disease classification model achieving 99.55% own-validation accuracy. – Implemented FastAPI REST APIs, PostgreSQL/PostGIS, Redis, Docker Compose, authentication, authorization, and object-level access control for secure deployment. – Integrated RAG, a Knowledge Graph, and LLM-based advisory generation with validation mechanisms to improve context-aware farmer recommendations. – Performed production validation with 464 tests passed, 0 failures, live backup/restore testing, and multi-service Docker deployment