varshith vishnu

Education

Annamacharya institute of science and technology rajampet - B. Tech, AI&ML

2022 – 2026

Rajampet · Grade: 8.56

SENTINEL CONSOLE — Multi-Agent AI Order Supervisor github.com/Vishnuvarshith91/Sentinel-Console.git

Architected a durable, hierarchical multi-agent system — a TinyLlama classifier paired with a
Qwen2.5 supervisor — orchestrated via Temporal workflows to autonomously monitor and resolve
order-lifecycle events end-to-end.
Engineered a deterministic rule-based fallback layer guaranteeing 100% demo uptime during LLM
outages, plus a live Next.js ops dashboard streaming workflow state, timelines, and agent memory in
real time.

CARDIODIAGNOSE — Cardiovascular Risk Prediction github.com/Vishnuvarshith91/CardioDiagnoise.git

Built a clinical ML pipeline on 12 patient-vital features; CatBoost and LightGBM hit 91.6%/92.4%
accuracy at 12ms/8ms inference — production-ready for real-time decision support.
Integrated SHAP explainability into a Streamlit dashboard, letting physicians trace risk scores to
individual biomarkers and boosting trust in simulated clinical trials.

HYBRID RAG SYSTEM — Document Question Answering github.com/Vishnuvarshith91/Hybrid-RAG-Model.git

Architected a hybrid retrieval pipeline (FAISS dense + BM25 sparse + neural re-ranking), cutting
hallucination rate ~30% vs. naive RAG on 50-page benchmarks.
Deployed a fully offline local-LLM backend via Ollama, achieving sub-2s end-to-end responses on an
8GB consumer machine.

YOLO POTHOLE DETECTION — Real-Time Road Defect Detection github.com/Vishnuvarshith91/YOLO26-Pothole-Detection.git

Fine-tuned YOLO on a 2,000-image custom dataset, reaching 0.7823 mAP at 30 FPS; shipped as a live
Streamlit app for road-maintenance teams.

MULTIMODAL SENTIMENT ANALYSIS github.com/Vishnuvarshith91/Sentimental-analyzer.git

Fused RoBERTa + VADER text sentiment with facial-expression CNN into one Flask REST API,
delivering real-time cross-modal emotion scores at 84% accuracy.

CardioDiagnose: A Custom Machine Learning Model for Real-Time Cardiac Risk Diagnosis Using CatBoost, LightGBM, and Deep Learning

Catalyst Convene - Kolkata with International Conference of Reminiscent Research on Artificial Intelligence (ICRRAI 2K25) · Sep 27 2025