About

Hey 👋, I’m Allen Philip!

ML engineer optimizing big models for real-time performance. Writing about inference, attention, and the edge of what’s practical.

About me

I’m a Machine Learning Engineer currently at Adobe, working at the intersection of large vision and audio models, LLMs, and GPU-level performance optimization. My work lives in the wild: production pipelines, low-latency inference systems, and GenAI features used by millions.

I specialize in making large models run fast and cheap—whether that means quantization, attention kernel optimization, or pushing TFLOPs on A100s without melting the cluster.

I started my career in backend systems and large-scale data engineering, eventually making the jump into ML systems in 2016. Since then, I’ve led and built pipelines for everything from retail demand forecasting to anomaly detection to Adobe Firefly’s generative stack.

This blog is my personal space to document what I’m learning, debugging, optimizing or just obsessing over.

What You’ll Find Here

Deep dives into ML inference and model optimization (FlashAttention, quantization, TensorRT, FSDP, etc.)
Notes and explainers on attention mechanisms, transformer internals, and how things really work under the hood
Reflections on books, engineering, and ideas that don’t fit into a Jira ticket