Rickston Pinto

About Rickston Pinto

I’m an AI Infrastructure Engineer focused on hardware acceleration, distributed…

Experience

1. Ai Software Engineer

   National Transfer Network

   Feb 2025 — Present

   Led the optimization of a large-scale recommendation engine by implementing a vLLM-based inference server architecture, achieving a 60% reduction in end-to-end P99 latency while building custom PagedAttention memory management components with CUDA/Metal kernel optimization- Implemented and benchmarked advanced speculative decoding techniques for production LLM deployment, achieving a 35% reduction in computational overhead while maintaining response quality through intelligent caching strategies that utilize transformer attention weights- Engineered a comprehensive benchmarking suite to track the performance of various inference runtimes (vLLM, TRT-LLM, SGLang), modeling production traffic patterns to identify and eliminate serving stack bottlenecks with automated stress testing capabilities- Published internal technical documentation on optimizing KV-cache handling for long-context models, improving memory efficiency across GPU clusters by 25% through advanced caching strategies and distributed memory coordination- Developed distributed inference optimization components, including tensor parallelism strategies, pipeline parallelism coordination, and semantic similarity-based request batching for multi-GPU serving architectures.

Education

• Birla Institute of Technology and Science, Pilani

  Bachelor of Engineering - BE, Electronics and Instrumentation Engineering

• Carnegie Mellon University

  Master's degree, Information Systems Management

  2024