Mihir Patel

Built an end-to-end bridge platform with async pipelines and a pluggable adapter architecture connecting enterprise VMS feeds (Nx Witness, Frigate) to Intel Live Video Captioning, enabling automated captions, alerts and bookmarking directly into the VMS without human intervention. Refactored backend microservices of Intel Open Edge Platform's video search and summarization for fault isolation and redesigned the React frontend into streamlined workflows, making the platform stable under concurrent usage with search results and summaries consistently accessible.

Implemented

Retrieval-Augmented Generation (RAG) with vector embeddings and NLP-driven query expansion to unify retrieval across video, audio and text modalities with a live evaluation system (MENLI + cosine similarity), and authored an internal research paper establishing a reproducible benchmarking framework for validating model improvements.

AI Software Development Engineer Intern - Intel Corporation - Bangalore, India

(2025-06)

• Built an end-to-end bridge platform with async pipelines and a pluggable adapter architecture connecting enterprise VMS feeds (Nx Witness, Frigate) to Intel Live Video Captioning, enabling automated captions, alerts and bookmarking directly into the VMS without human intervention.
• Refactored backend microservices of Intel Open Edge Platform's video search and summarization for fault isolation and redesigned the React frontend into streamlined workflows, making the platform stable under concurrent usage with search results and summaries consistently accessible.
• Implemented Retrieval-Augmented Generation (RAG) with vector embeddings and NLP-driven query expansion to unify retrieval across video, audio and text modalities with a live evaluation system (MENLI + cosine similarity), and authored an internal research paper establishing a reproducible benchmarking framework for validating model improvements.

Assistant System Engineer - Tata Consultancy Services - Gandhinagar, India

(2023-11 - 2024-08)

• Developed RESTful microservices using FastAPI with async endpoints and Pydantic validation for an internal workflow automation platform, reducing cross-team integration time from days to hours.
• Dockerized backend services with multi-stage builds and Docker Compose, enabling reproducible one-command deployments across dev, QA and production environments and eliminating environment drift.
• Profiled and optimized an enterprise reporting service pipeline by restructuring database queries and introducing async batch processing in FastAPI, significantly improving API throughput and response times under load.

M.Tech - Computer Science - Sardar Vallabhbhai National Institute of Technology Surat (SVNIT) (2024 - 2026)

B.E - Computer Science - The Maharaja Sayajirao University of Baroda (2019 - 2023)