Aayush Garg

Notes on applied ML engineering, LLMs, and building AI products.

I am an applied machine learning engineer focused on building AI-driven products and systems people can rely on. At Jiffy, I build image enhancement pipelines, deploy ML models at scale, and develop LLM/VLM applications across product verticals.

💼 Current Work

I am currently building an in-house AI design platform, developing evaluation and prompt-optimization workflows, and exploring applied LLM fine-tuning for structured data extraction. My background is in computational geophysics, which keeps me grounded in first principles and pragmatic experimentation.

If you’re interested in my work or want to collaborate, feel free to reach out via email or connect on LinkedIn.

📮 Blog

I write about practical ML engineering, LLMs, and lessons learned from building AI products. Below are my recent posts.

Date	Title
12/30/25	Deriving the DPO Loss from First Principles
12/25/25	Deriving the PPO Loss from First Principles
12/3/25	What I Learned Building SFT from the Ground Up
9/10/25	A Guide to Building Custom Nodes in ComfyUI
9/8/25	Building GPT from Scratch: Following Karpathy’s Tutorial
9/2/25	Key Takeaways from Lecture 1: LLM Evaluation Lifecycle
7/15/24	Part III: Fine-tuning Llama-3-8B for Structured Functional Representation Extraction
7/9/24	Part II: Comparison of Model Performances on Structured Functional Representation Extraction
7/3/24	Part I: Baseline Evaluation of GPT-4o for Functional Representation Extraction
6/30/24	Step-by-Step Guide to Setup Your Personal GPU Server

📬 Reach Out

The best way to follow my work is on GitHub, Twitter, or LinkedIn.

Template inspired by hamel.dev and Lil’Log.