Harshith Reddy Thammineni

WORK HISTORY

EDUCATION

ABOUT HARSHITH REDDY THAMMINENI

My work spans end-to-end compiler design, including graph-level transformations, quantization-aware optimizations, and low-level code generation using LLVM and custom runtime abstractions. I specialize in building compilation pipelines that target embedded and high-throughput systems, balancing model accuracy with system constraints.Previously, my research contributions included the publication of ELEGAN: An Efficient Low Light Enhancement GAN for Unpaired Supervision, a lightweight, unsupervised image enhancement model based on attention-guided generative adversarial networks. It introduced a self-regularized illumination attention map and a modified residual dense block to deliver state-of-the-art results in low-light enhancement.My technical skills span a broad ecosystem, including Python, C++, Bash, LaTeX, and SQL, with deep experience in PyTorch, TensorFlow, LLVM, ONNX, Scikit-learn, and GIT. I have also worked extensively with cloud platforms such as AWS and Google Cloud, and databases including MongoDB and Snowflake.Driven by a passion for system-level ML optimization and real-world deployment, I aim to push the boundaries of compiler design, performance quantization, and embedded inference. Feel free to connect with me on LinkedIn or reach out at h•••@illinois.edu to explore potential collaborations or opportunities.