01
PDF RAG Pipeline
A multimodal retrieval-augmented generation pipeline for exploring ArXiv research papers through natural-language questions and optional image queries.
Software Engineering · Machine Learning · Systems & DevOps Engineering
I’m Vishal, a Software Engineer with a background in Distributed Systems, Backend Development, DevOps, and Applied AI. I enjoy designing reliable, production-ready platforms from cloud infrastructure and data pipelines to machine learning and computer vision systems that solve real-world problems. After earning my Master’s in Computer Science from Arizona State University, I’m focused on building scalable software systems with an emphasis on performance, maintainability, and clean engineering.
Seattle, WA
Current focus
Building AI-powered software platforms for intelligent energy systems at ASU’s Laboratory for Energy and Power Solutions (LEAPS), with a focus on backend engineering, multi-agent workflows, cloud infrastructure, and scalable data pipelines.
Selected projects
Personal and academic projects spanning AI, data, energy systems, computer vision, and full-stack development.
01
A multimodal retrieval-augmented generation pipeline for exploring ArXiv research papers through natural-language questions and optional image queries.
02
A graph-based streaming analytics system that models NYC Yellow Taxi trips in Neo4j, runs PageRank and breadth-first search, and extends batch ingestion through Kafka and Kubernetes.
Experience
Selected roles spanning intelligent energy, edge computer vision, data processing, and digital solutions.
Jun 2026 — Present
Current
ASU Laboratory for Energy and Power Solutions (LEAPS)
Mesa, AZ
Building AI-powered software platforms for intelligent energy systems by developing scalable backend services, cloud-native applications, multi-agent workflows, and data pipelines that support real-world energy optimization and decision-making.
Dec 2023 — Jun 2024
Eternal Robotics
Hyderabad, Telangana, India
Built real-time computer vision inspection systems for industrial manufacturing. Redesigned a tractor-engine component inspection pipeline to combine object detection with location-based classification, improving robustness to reflections and metallic surfaces and achieving over 95% inspection accuracy. Migrated inference to NVIDIA DeepStream and TensorRT on Jetson edge devices, reducing latency by approximately 85%, and integrated real-time PLC data ingestion, configuration mapping, validation, and OK/NG feedback to support multiple product variants without manual intervention.
May 2023 — Sep 2023
Worley
Hyderabad, Telangana, India
Developed a computer vision pipeline to identify changes between revised engineering PDF drawings despite variations in position, scale, and orientation. Experimentally compared ORB with brute-force matching against SIFT with KNN matching and Lowe's ratio test, documented the speed and reliability tradeoffs, and helped select the SIFT-based approach, which achieved approximately 98% matching accuracy on unseen data.