About Me
Hi! I’m Pouya (Reza) Rostam. I’m an Applied Scientist at Amazon working on LLMs and ML systems for Alexa.
💼 Experience
Amazon — Applied Scientist (2024 - Present)
Pioneered CI/CD paradigm for LLM development, reducing dev-to-prod cycle by weeks. Developed automated prompt engineering framework enabling optimization across 8 languages. Built agentic system for automated analysis of customer interactions, reducing investigation time from days to hours.
Picovoice — Senior Deep Learning Researcher (2022 - 2024)
Led CUDA kernel development for picoLLM inference engine, achieving 3x speedup. Led end-to-end development of Eagle (speaker recognition) and Falcon (diarization) engines from concept to market. Led TTS research team of 3.
Picovoice — Software Engineer & Embedded Lead (2020 - 2022)
Ported voice AI to MCUs and WebAssembly, cutting resource consumption by 40%. Built Shepherd, a no-code platform for embedded voice AI. Architected CI/CD that reduced release cycles from weeks to days.
McMaster Manufacturing Research Institute — Instructor (2022)
Designed and delivered modules in Python & MATLAB programming, earning 90% student satisfaction.
UBC — Graduate Research Assistant (2016 - 2020)
Researched hybrid Gaussian process methods for robust model predictive control at the Control Engineering Laboratory. Managed an industrial research partnership and mentored 6 students.
🎓 Education
Ph.D. Control Systems — University of British Columbia
Thesis: A Hybrid Gaussian Process Approach to Robust Economic Model Predictive Control
M.Sc. Mechanical Engineering — Sharif University of Technology
Thesis: Control of Adaptive Optics Systems Using Transverse Actuators
🔬 Expertise
NLP & AI
Speech Processing · Transformers · RAG · Text Classification · Information Extraction · LLMs · Fine-tuning
Frameworks & Libraries
PyTorch · LangChain · LangGraph · Hugging Face · NumPy · Pandas
Performance & Systems
CUDA · C/C++ · Rust · Assembly · WebAssembly
Cloud & MLOps
AWS · CloudFormation · Serverless · Docker · Kubernetes · CI/CD