SINCE 2026
ML Performance Engineer
- Full-Stack-to-Kernel Optimization
- Frontier-Model Performance on TPU & GPU
- On-Device & Open-Model Efficiency
- Applied AI with Measured Impact
ML Performance Engineer · Google · San José State
Anudeep Goud
San Jose, California
The performance engineer who chased the bottleneck from headroom analysis down to frontier-model kernels.
Anudeep Goud is an ML Performance Engineer at Google, where he tunes frontier models above 300B parameters on the XLA compiler, writing kernels in Pallas, CUDA, and Mosaic for TPUs and NVIDIA GPUs alongside DeepMind. He reached the kernel layer by descending it:, LLaMA 3.1 on laptop NPUs at HP, storage-and-inference systems at NetApp. Back in Hyderabad, he also ran the animal-welfare wing of Street Cause.
- >300BParams in frontier models he tunes at Google
HAPPENING NOW
STANDING TODAY
HOW THEY THINK4 BELIEFS
Four beliefs explain a steady descent toward the hardware.
THE IMPACT
Eight roles, one thread — make the computation faster and prove it.
THE TIMELINE
One instinct — find the bottleneck, then move a layer lower — from SQL scripts to frontier-model kernels.
The whole record, newest first – tap any moment and Aster tells the story.
2026–PRESENT
Down to the kernel: frontier performance at Google
As an ML Performance Engineer in Google's CoreML Frameworks and Compilers group, Anudeep tunes models above 300B parameters on the XLA compiler, writing kernels in Pallas, CUDA, and Mosaic across TPU generations Sunfish, Ghostfish, and Jellyfish and NVIDIA's GB200 and GB300. He breaks bottlenecks in Google's Gemini and Gemma and third-party models like Qwen, DeepSeek, and Kimi, working with DeepMind.
2025–2026
The systems bridge at NetApp
As a Member of Technical Staff II at NetApp in San Jose, Anudeep worked where distributed storage meets LLM inference. The role paired Linux-kernel, KVM, RDMA, and NFS/DFS systems with PyTorch, VLLM, CUDA, and quantization — the plumbing that decides whether accelerators stay fed. It is the pivot from his applied-GenAI phase into pure performance engineering at Google.
2023–2025
Grad school and the pivot into GenAI
Moving to California for an MS at San José State, Anudeep reframed his career around LLMs. As a GenAI Graduate Research Assistant he deployed and scaled custom LLMs on AWS for the biomedical department; internships at HP (LLaMA 3.1 on NPUs) and EXL (implementing research papers) followed in quick succession, and he mentored 40+ peers as a Statistics TA.
“Fostered a deeper understanding of statistical methods by mentoring 40+ peers, leveraging expertise in Quantitative Analysis, linear and logistic regression, significance testing, and statistical modeling”
2018–2023
Hyderabad: first code, first optimizations, first service
In Hyderabad, while earning his BTech at JNTU, Anudeep ran three student community organizations before his technical career took off.
“Optimized SQL scripts to enhance data loading efficiency… ultimately providing Tableau users with 40% faster access to updated data.”