Backend ML Systems & RAG Engineer

Meghkumar Patel

San Jose, CA Envelope **************@*****.*** j Phone-Alt 669-***-**** j LINKEDIN meghkumar-patel j Github meghp73 EDUCATION

San Jose State University Aug 2024 – May 2026

Master of Science in Software Engineering San Jose, CA Relevant Coursework: Machine Learning, Computer Architecture, Compiler Design, Operating Systems, Distributed Systems Nirma University Aug 2020 – June 2024

Bachelor of Technology in Computer Science and Engineering Ahmedabad, India SKILLS

Languages : C++, Python, CUDA C++, Java, Go, TypeScript, SQL AI / ML : PyTorch, Deep Learning, LLMs, RAG, AI Agents GPU & Compiler : CUDA, LLVM, Clang LibTooling, AST Matchers, Compiler Design, Static Analysis Systems : Linux, Parallel Programming, Multithreading, Concurrent Programming, Distributed Systems Cloud / DevOps : AWS, Docker, Kubernetes, Terraform, CI/CD Databases : PostgreSQL, MySQL, MongoDB, Redis, ChromaDB EXPERIENCE

Parshwanth Formatting (OPC) Pvt. Ltd. Jan 2024 – Jul 2024 Software Engineer Vadodara, India

Developed high-performance backend components in C++, optimizing concurrency, memory usage, and latency.

Improved service observability with structured logging, error tracking, and real-time runtime debugging, enabling faster incident response and reducing MTTR for production issues.

Defined clear REST service contracts between systems, ensuring seamless cross-functional integration and enabling frontend, QA, and partner teams to ship faster with fewer rollbacks.

Collaborated on software architecture and performance reviews focused on scalability, concurrency, latency reduction, and system reliability.

Prolog Infotech Jun 2023 – Aug 2023

Backend Developer Intern Rajkot, India

Engineered backend services in Node.js/Express.js with caching strategies and query optimization that reduced API re- sponse times by 20%, improving reliability for millions of requests.

Built a production-grade observabilitystackapplication logging, error tracking, and bottleneck profiling across load-balanced service instances that reduced debugging time and improved fault isolation in high-throughput environments.

Collaborated with QA and frontend engineers to improve test coverage and resolve integration failures, driving faster, more stable release cycles aligned with an agile delivery model. PROJECTS

CUDA Kernel Performance Analyzer (LLVM/Clang LibTooling) Github LLVM, CUDA, Compiler Tooling, Static Analysis

Built a compiler-assisted CUDA kernel analysis framework in C++ using LLVM/Clang LibTooling to identify GPU perfor- mance bottlenecks and correctness issues before runtime.

Developed compiler analysis passes over Clang ASTs to analyze thread indexing, shared memory utilization, synchro- nization patterns, occupancy-related launch configurations, and kernel execution characteristics.

Designed scalable source analysis pipelines supporting large CUDA codebases with automated validation suites for kernel parsing, launch analysis, and diagnostic generation.

Researched CUDA optimization techniques including memory coalescing, shared memory utilization, synchronization overhead, and warp-level execution to guide kernel optimization recommendations. TaskMaster – Multithreaded Task Scheduler Github C++, Multithreading, Synchronization, STL

Designed a lock-efficient multithreaded task scheduler in modern C++ using concurrent priority queues, condition vari- ables, and worker pools to maximize CPU utilization under parallel workloads.

Implemented asynchronous producer-consumer logging pipelines that eliminated blocking I/O and minimized synchro- nization overhead during high-throughput execution.

Implemented thread-safe task lifecycle management with retry and exponential backoff, improving scheduler robustness under transient failures.

DataTrust – Policy-Aware Enterprise AI Assistant with Secure RAG Github Python, FastAPI, ChromaDB, RAG

Built a production-grade AI assistant with multi-step RAG, agent reasoning, token-aware chunking, and embedding-based retrieval backed by ChromaDB designed to scale reliably across concurrent enterprise users.

Developed FastAPI-based backend services for document ingestion, retrieval orchestration, authentication, and audit logging, enabling scalable concurrent AI inference workflows with end-to-end observability.

Contact this candidate