Sohail S.

Compiler Engineering (LLVM) ML / Deep Learning Compiler Development Machine Learning Engineering C++ Development Python Development (6+ yrs) JavaScript Development CPU, GPU perf optimisation C++ ==== * Commercial C++ experience: 3-4 years * Non-commercial C++: 6+ years * C++14, C++17, C++20 (modules), C++23 (limited). * Rust (3months) CMake, STL, Boost, GTest, gcov C++ modern multi-threaded, stdexec C: Embedded C (embedded Arm) Robotics: ROS, TF2 Emscripten (frontend C/C++ for browser jasm & wasm) Interest in: Zig, Jai, Rust. Compiler Engineer ============== LLVM, MLIR, C4ML Deep Learning & Machine Learning =========== TensorFlow 2,1, PyTorch, ICA, Bayesian model selection, Neural Networks (from Scratch), ResNet, Transformers, Contrastive State Space modelling (Kalman), Numerical Stochastic process modelling, Ito. Point Processes, Gaussian Processes. Statistical models, small sample size bias, AIC/BIC, Info. Theoretic Predictive Processing & Helmholtz Machines (theoretical), Bayesian (theoretical), RL (theoretical) GPU ==== CUDA, PTX, LLVM on GPU, MLIR, OpenGL, glsl Random number generation (GNG) on GPU & CPU Python ====== * Python developer 6+ years (concurrent, async & quantitative, Numpy (extensive), REST, Flask, Matplotlib (extensive) animations, pygame, low-level Linux: python/C), C++ binding, Types, Numba, mojo (MLIR) JavaScript ========= * JavaScript: 6 years * Front-end (native, performance-optimised javascript (extensively), React.js, Redux, Angular) * Backend: Node.js (extensive), Web APIs, API gateways (reverse proxy), OAuth, Open Data, Open Banking, testing frameworks (extensive: cucumber, chai, mocha/karma, Sinon, etc), pm2, etc * JavaScript: ( async, promises, etc), WebGL (full OpenGL program), three.js (extensive), Emscripten (extensive), AngularJS1/2, web sockets, web workers (++), PWA, Jasmine with Karma for TDD/BDD, web assembly, TypeScript (limited), GraphQL (limited), memory and performance profiling (back-end and front-end), more Java ===== * Java programmer since 1998 (early adopter) plus teaching Java in various companies * JVM internals (wrote a JVM interpreter PoC) * Sprint/Boot (limited) Core Coding: ======= * Algorithm design IOI/ACM * Very skilled at complex refactoring & modernising legacy code (without breaking) * Unparalleled Debugging skills * Excellent test skills * Programmer since age 11 * Low-level programming (incl. interfacing between custom-built Hardware with C/C++, assembly ) * HDL: VHDL, AmaranthHDL, open source HDL, HLS. * FPGA, AWS F2 (cloud FPGA for heterogeneous) * GPU using CUDA for NVidia (A100: 2024, K40: 2013) * Linux systems programming (wrote Docker from scratch) * Experience in distributed prog. (Job scheduler for cloud, robust protocol design). *DevOps/Infrastructure: Docker, CI, github Actions workflows, Jenkins & Groovy, Terraform, F2. * I am familiar with the inner workings of cloud technologies such as Docker (wrote an equivalent of Docker). * Designed custom standalone hardware from scratch (direct VHDL on FPGA) Computer Vision ============= GAN (Generative Adversarial Networks) implementation from scratch on TF1 Tensorflow & PyTorch for vision OpenCV Optical flow (nu-Optic Flow), Mathematical Programming & Applied Mathematics ============= Linear Algebra & Numerical Linear Algebra (extensively) & Sparse Matrices. Non-linear constrained Optimisation (Lagrange multipliers, KKT) Parametrised probability distributions (ICA, circular distributions), Copulas (A. Genz Integration algorithm for Gaussian CDF), NCE MCMC (Metropolis-Hastings), Stochastic processes (Euler–Maruyama SDE), Point processes (inhomogeneous Poisson + more), Physics simulations (robotics, biophysical/neural, optics/rendering, Quantum ), Expectation-Maximisation/EM/MM (for Hinton-Ghahramani for Kalman), Data Science & Stat. ============ I have a PhD in computational Neuroscience. Very good theoretical knowledge of Machine Learning & Neural Networks, including Statistical ML and Deep Learning, Information Theory. Numerical linear algebra, high-performance numerical calculations, etc. I have used a variety of Statistical modelling techniques (using Bayesian approaches), hands-on experience with statistical estimations, Entropy estimation. 3D programming ============= SDF modelling, Geometric modelling, 3D Shape modelling, Mesh geometry (subdivision, adaptive: performant & vectorized computational geometry algorithms from scratch), polygonization of implicit /algebraic surfaces, Marching Cubes, Sphere Tracing/Ray Marching. Interest in: NeRF, Gaussian Splatting. Computer Graphics (extensive). Robotics ======= ROS2 (extensive), DDS, Localisation (bearing-only SLAM), VPR, Optic flow (limited) Audio programming ================ Created products for: Realtime Audio. SPDIF protocol (VHDL-implemented & built hardware from scratch). VST plugin development, Direct Sound. Digital Filter Design & phase, FFT (extensive), Wavelets. Physical sound modelling, Granular modelling.