Shreyas Sangalad
AI Engineer with hands-on experience architecting production multi-agent LLM systems (Lang Graph, RAG, Vertex AI) and full-stack applications on GCP. Published IEEE researcher in deep learning (94.2% accuracy, Efficient Net-B3). Strong foundation in Java, Python, and cloud-native backend development. Seeking a full-time AI Engineer role.
Education
SJB Institute of Technology (SJBIT)
B.E.
CGPA: 6.8. Coursework: DSA, DBMS, Web Technologies, Computer Networks, OS, Cloud Computing.
Skills
JavaPythonSQLCC++LangChainLangGraphGemini LLMRAGMulti-Agent OrchestrationPyTorchVertex AICloud RunCloud StorageCloudDocument AIPub/SubIAMREST APIsFlaskWebRTCJDBCServletJSP
Projects
Multi-Task AI Retinal Screening System
Machine learning
Designed a multi-task deep learning framework using EfficientNet-B3 with dynamic loss balancing for simultaneous retinal disease classification and lesion analysis, achieving 94.2% accuracy. Integrated Grad-CAM explainability for clinical interpretability; published at IEEE IC-ICNS 2026.
Nexus — Autonomous AI Customer Support Agent
Agentic ai
Built a production-grade ReAct agent (LangGraph + Gemini) that autonomously routes between a RAG pipeline (ChromaDB) and live order lookup tools to resolve customer queries end-to-end. Implemented real-time SSE token streaming via FastAPI, persistent multi-turn memory with AsyncSqlite-Saver, human escalation detection, and dynamic multi-language response injection.
Publications
Multi-Task EfficientNet-B3 With Dynamic Loss Balancing for Retinal Disease Detection and Explainability
Shreyas Sangalad, Vijayalakshmi B., Varun S., Sharan V. Neelgal, Srujan H. S.
IEEE IC-ICNS 2026. DOI: 10.1109/IC-ICNS68863.2026