Kishan Teja Repaka

About Kishan Teja Repaka

Hey there — I’m Kishan, a Data Scientist and AI Engineer who loves turning complex machine learning ideas into real, working systems that make life (and work) a little smarter.Over the past few years, I’ve built everything from Generative AI platforms and retrieval-augmented generation (RAG) systems to LLM-based agents that can reason, adapt, and even collaborate. My journey has taken me through roles at Amazon, Navy Federal Credit Union, and Maersk, where I’ve worked on projects involving distributed ML, document intelligence, and cloud-scale deployment — basically, the messy, exciting intersection where data meets creativity.These days, I spend most of my time fine-tuning and optimizing Large Language Models (LLMs) — experimenting with instruction tuning, LoRA/QLoRA, and RLHF, and building systems that can actually think through problems instead of just predicting the next word. I’ve tinkered with frameworks like LangChain, DeepSpeed, and vLLM, and I’m fascinated by how modular, multi-agent AI systems might evolve into the next big leap in reasoning.On the engineering side, I love scaling AI pipelines across AWS, Azure, and GCP, using tools like PyTorch, FastAPI, Hugging Face, TensorRT, and NVIDIA Triton. Whether it’s speeding up training by 70%, improving retrieval accuracy, or deploying production-ready agents — I’m happiest when building systems that actually work at scale.But underneath all the acronyms and models, I’m just a curious problem-solver. I enjoy exploring new research papers, connecting ideas across domains, and asking “what if we tried it this way?” more times than my teammates probably appreciate.I’m especially passionate about LLM research, generative reasoning, and agentic architectures — where AI stops being just a tool and starts becoming a partner in creativity and problem-solving.If you’re building something exciting in LLMs, RAG, fine-tuning, or AI systems that can collaborate and learn, let’s connect! Always down to talk about new ideas, cutting-edge experiments, or the occasional wild hypothesis about where AI is headed next.

Experience

1. Software Development Engineer

   Amazon

   Jul 2025 — Present · Seattle, WA, US

   Led JDK, Hibernate, and Spring migration for Yoshi Asset Service, Prime Video’s core platform managing video encoding, metadata processing, and content lifecycle workflows across 7 services and 67 packages. Deployed staged pipeline, maintaining 100% stability across 2,700+ unit tests with zero production regression.• Architected and implemented S3 streaming solution to replace legacy S3 Copy operations in ATV Replication Service, reducing memory footprint and improving transfer reliability through chunk-based processing. Implemented circuit breaker patterns with automatic fallback, and executed phased rollout (0%→100%) with monitoring.

Education

• National Institute of Technology, Kurukshetra, Haryana

  Bachelor of Technology, Computer Engineering

• Arizona State University

  Master of Science - MS, Computer Science