Machine Learning Engineer

@farooq-azam-khan

Specializing in

delivering scalable AI solutions.

with VLLMs, Ollama, and NVIDIA TensorRT for optimized model serving pipelines.

interested in

graph algorithms

reinforcement learning

Enhanced LLM Agents

Bridging best parts of theory and practice to create impactful, intelligent systems.

📰 Fresh From the Lab

Maximum Likelihood Estimation in ML

Crunch numbers like a pro—Master MLE & Gradient Descent.

Scaling LLMs with Triton

Deploy LLMs at scale with NVIDIA Triton Inference Server.

TF-IDF Demystified

Unlock text insights—Master Term Frequency & Inverse Document Frequency.

Scalable Vector Comparison

Compare massive vector sets efficiently—A hands-on guide.

Cosine Similarity 101

Master text vector similarity with NumPy & PyTorch in minutes.

🤖 Experiments & Creations

Race Car RL Agent

Developed a Deep Q-Learning agent in a simulated racing environment (e.g., OpenAI Gym CarRacing) to autonomously navigate tracks by learning from reward feedback.

Read the Paper →

Garbage Classification with CNN

Transfer-learned an EfficientNet backbone to classify multiple waste categories in images, enhancing model performance through data augmentation and fine-tuning techniques.

Read the Paper →

Haiku Generator

Explored prompt-engineering strategies with large language models to generate structured haikus, enforcing syllable constraints and thematic coherence.

Source Code →

D3.js Reference Tutorial

Interactive web guide demonstrating core D3.js features and visualization techniques.

Source Code → Live Demo →

Chat App

Real-time chat app with WebSocket messaging; Elm frontend and Express.js backend.

Source Code →

Twitch Python Discord Bot

Python Discord bot with Twitch API integration for live stream alerts and moderation.

Source Code →

Rust Tauri App

Cross-platform desktop color picker built with Rust and Tauri, showcasing native UI integration.

Source Code →

Guided Sudoku Solver

Interactive Sudoku solver with step-by-step guidance built in Python using a backtracking algorithm and heuristic improvements.

Source Code →

Roadmap

APPLIED REINFORCEMENT LEARNING

Reading: Reinforcement Learning (Sutton & Barto), Deep Learning (Goodfellow et al.)

Building with: Markdown Python Obsidian Pytorch Numpy

Exploring: RL Algorithms

HIGH-PERFORMANCE SYSTEMS

Reading: Rust In Action (Tim McNamara )

Building with: TensorRT Triton CUDA Postgres DB

Exploring: Convex, Distributed Systems

Design Philosophy:

Prototyping v0/v1 solutions and analyzing data with Python, leveraging its rich ML ecosystem.
Optimizing for speed and memory safety in v2 systems using Rust. Exploring GPU acceleration via CUDA and Triton.
Building robust, full-stack applications with TypeScript for maintainable web code and Rust for performant back-end logic.