Research & Engineering

Blog

Technical deep dives into AI infrastructure, neural network compilation, and the systems we're building at Dheyo.

Latest

reinforcement-learningtext-to-image / Mar 15, 2025

Enhancing Text-to-Image Prompts using Reinforcement Learning

How GRPO-based reinforcement learning improves prompt enhancement for text-to-image models while preserving logical and visual constraints.

Read article

All posts

Mar 10, 2025 llm-inference energy-efficiency

Evaluating Energy Efficiency and Performance in LLM Inference and Introducing DheyoGPUpd

A study of tokens-per-joule trade-offs in LLaMA-3.1-8B inference and an introduction to DheyoGPUpd for unified GPU telemetry and power control across heterogeneous NVIDIA and AMD environments.