Hasan Ahmed

Math

• Linear Algebra
• Differential and Multivariable Calculus
• Statistics & Probability
• Hypothesis & A/B testing (t-test, chi-square, Mann-Whitney etc)
• Correlation and covariance (Pearson's, etc)
• Multivariate analysis
• Bootstrapping methods (sampling, etc)
• Power & meta-analysis (Confidence intervals, Margins of error, etc)
• Bayesian vs Frequentist statistics

ML Hardware Skills

• Customizing CUDA kernels, batching, and processing
• Extending the cuDNN library for lower-level control
• Solid grasp of architectural designs in Hopper, Ampere, and Blackwell Nvidia GPU architectures
• An understanding of LLM inferene engines such as vLLM and TensorRT-LLM, how they work, and their limitations
• Writing custom extensions for TensorRT-LLM to use via Graph Rewriting API
• PagedAttention, OpenAI-Style endpoints, KV caching, and other features inference engines provide

Deep Learning Skills

• PyTorch, Keras, Tensorflow
• Neural networks and the available architectures (RNNs, LSTMs, CNNs, Transformers, Vision Transformers, etc)
• Deep learning practices in regard to learning rates, optimizers, regularization, batching, and overall training expertise
• Understanding of underlying concepts such as chain-rule & backpropagation, gradient descent, exploding/vanishing gradients, and more
• A good foundation of underlying mathematical concepts in Linear Algebra, Calculus, and Stats/Probability
• The ability to read research paper and implement the steps to reproduce and verify, as well as use exisiting work as inspiration to continue researching
• A skill for creativity in regard to LLM bottlenecks such as attention mechanisms, context extension, and throughput from studying models such as LongFormer & BigBird
• Deploying and monitoring models in production via pipelines using MLFlow or Kubeflow

Data Science Specific

• Power BI, Excel, Powerpoint
• Programming languages such as Python and SQL
• Numpy, Pandas, Matplotlib, Seaborn, and Tableau
• Machine learning algorithmic EDA via Scikit-learn (KNN, Isolation Forests, Decision Trees, Linear & Logistic Regression, and how to verify metrics)

Software engineering

• Programming languages: JavaScript, Python, SQL, Java, C, Ruby and markdown languages such as HTML and CSS
• Frontend: React 18 (client and server components, streaming, etc), NextJS, Web API's, Redux, TailwindCSS, and extensive technical knowledge on web browsers (cookies, sessions, rendering, etc)
• Backend: NodeJS, Express, PassportJS, Microservices, API's, Server deployment
• Other: modular design patterns, testing, workflows, best practices, monitoring, etc