Marvin

^,-#?_<-/

MEEnG

Electrical Engineering

AI, DSP & GPUs

about me

Masters Electrical and Electronic Engineering at UMass Dartmouth, research focused on Signal Processing, Bayesian Statistics, and Beamforming, depth spanning numerical computing and probabilistic search algorithms.
Topic of research: probabilistic active-sensing search algorithms (greedy vs entropy-driven) in Monte Carlo simulations and robotic experiments, comparing algorithmic duration performance.

Active work focuses on Large Language Models, Efficient Computing and Deep Learning, with emphasis on efficient inference, distributed systems, and edge deployment.

I've previously built performance-critical ML systems, including lightweight inference engines for state-of-the-art small models, and contributed technical writing on device-level parallelism, distributed training (DDP/FSDP), and scalable model architectures.

Thankful to amazing mentors; Dr John R. Buck (graduate advisor, global lead expert on Signal Processing), Dr Ana Doblas (Optical Systems Expert, enjoyed her classes so much), Dr Kiragu Henry (Undergraduate advisor and mentor)!

work experience

selected publications

Impact of Beamwidth on Infotaxis Search Strategy Performance

John R. Buck, and Marvin Mboya

Signal Processing in Acoustics, 2025

Acoustic Society of America

latest blogs

Continuous Batching & Dynamic Scheduling

Implemented LFM's Small Language Model, explaining architecture concepts guided by pioneering research papers, optimized inference with hybrid caching, batched decoding, dynamic scheduling, ragged prefill, achieving major decoding speedups.

Jan 16th, 2026

Transformers to Vision Transformers: Scaling across GPUs

Scaling Vision Transformers: From Transformers to ViTs, exploring embeddings, attention, and GPU optimization. A hands-on journey from NLP roots to vision models, with PyTorch code, efficiency tricks, and inference pipelines.

July 5th, 2025

Neuron, the (to-be) plugin for Flutter

Neuron is a Flutter plugin-in-progress for ML, unifying TensorFlow & PyTorch with preprocessing, inference, and postprocessing. Built with MethodChannels and open-source for community contribution, it aims to simplify model deployment in apps.

June 7th, 2023

Asynchronous Programming in Julia

An intro to asynchronous programming in Julia using Tasks, macros, and Channels. Learn scheduling, @async, producer-consumer patterns, and binding tasks/channels—building a foundation for parallel and concurrent computing.

March 31st, 2023

latest courses

From Tensors to Residual Learning

Marvin Mboya

Calculus, Deep Learning (PyTorch), and Research

This in-depth course dives from tensors and calculus to graph-based differentiation, explores ResNet papers, implements them step by step, and finally imports pretrained ResNets in PyTorch for real-world inference.

July 20th, 2024

Learning Vue: Building Scientific Calculator

Marvin Mboya

Web-App Dev Dive

This in-depth course introduces Vue, a lightweight JS framework for reactive, declarative UIs. You'll master its core concepts by directly building a scientific Google-style calculator with the Composition API, step by step.

April 25th, 2023

Learning Flutter

Marvin Mboya

Mobile-App Dev Dive

Deep Dive into Flutter, built on Dart, a cross-platform UI toolkit. With hot-reload, plugins, and a layered architecture (App, Framework, Engine, Embedder, Runner), it lets you build fast, native-like apps from a single codebase.

April 8th, 2023

Basics Tour

Marvin Mboya

Procedural, Functional & OOP Dive

Hands-on course covering Julia (parallel/ML), Dart (mobile), Go (CLI/cloud), Python (ML), JavaScript (Web). Learn core computing concepts beyond syntax, from logic to advanced structures, to adapt confidently across diverse programming environments.

March 21st, 2023

0.00