Brian Mwangi | AI Engineer

Journey Through Neural Networks and Innovation

I'm Brian Mwangi, and this is the story of my relentless exploration into the depths of artificial intelligence. From optimizing neural networks with Unsloth to pushing CUDA to its limits, every line of code tells a story of curiosity, breakthrough moments, and the pure joy of discovery.

Deep Learning Explorer

Pushing the boundaries of neural architecture design and optimization

CUDA Innovator

Crafting custom kernels and maximizing GPU performance for AI workloads

RL Pioneer

Designing intelligent agents that learn, adapt, and surprise even me

Brian Mwangi - AI Explorer

The Journey Continues

This isn't just a portfolio, it's a living documentation of my adventure through the ever-evolving landscape of AI. Each project represents a milestone, each breakthrough a new chapter in understanding what machines can learn and how we can push them further.

Unsloth OptimizationTensorRT DeploymentReinforcement LearningNeural Architecture SearchCUDA Programming

About Me

Passionate AI Engineer with expertise in optimization, deployment, and research. I specialize in making AI models faster, more efficient, and production-ready.

AI Optimization

Expert in optimizing AI models for production deployment

UnslothModel CompressionQuantizationPruning

GPU Computing

Advanced GPU acceleration and inference optimization

CUDATensorRTcuDNNNVIDIA Triton

AI Research

Cutting-edge research in AI algorithms and architectures

Reinforcement LearningDeep LearningNeural Architecture Search

MLOps

End-to-end machine learning operations and deployment

Model DeploymentMonitoringAutoMLCI/CD

Featured Projects

A showcase of my work in AI optimization, GPU computing, and cutting-edge research projects.

Optimization
Unsloth Model Optimization
Optimized large language models using Unsloth, achieving 5x faster training speeds with 80% memory reduction.
UnslothPyTorchCUDALoRA
✓ 5x Training Speed
✓ 80% Memory Reduction
✓ Production Ready
GPU Computing
CUDA-Accelerated Neural Networks
Custom CUDA kernels for neural network operations, resulting in 3x performance improvements over standard implementations.
CUDAC++cuDNNTensorRT
✓ 3x Performance Boost
✓ Custom Kernels
✓ Low Latency
RL
Reinforcement Learning Trading Bot
Deep RL agent for algorithmic trading using PPO and advanced market analysis techniques.
PPOStable-Baselines3GymNumPy
✓ 23% Annual Return
✓ Risk Management
✓ Real-time Trading
Deployment
TensorRT Model Deployment
End-to-end deployment pipeline converting PyTorch models to optimized TensorRT engines for production inference.
TensorRTONNXDockerKubernetes
✓ 10x Inference Speed
✓ Scalable
✓ Cloud Ready
RL
Multi-Agent RL Environment
Complex multi-agent reinforcement learning system with custom environments and reward shaping.
Multi-Agent RLRay RLlibOpenAI GymPython
✓ Emergent Behavior
✓ Scalable Architecture
✓ Research Impact
Research
Neural Architecture Search
Automated neural architecture search system that discovers optimal model structures for specific tasks.
AutoMLPyTorchOptunaCUDA
✓ Automated Discovery
✓ SOTA Results
✓ Efficient Search

Chronicles of an AI Explorer

Welcome to my digital journal—where I document the highs, lows, and eureka moments of my AI adventure. These aren't just technical posts; they're stories of discovery, failure, breakthrough, and the pure joy of pushing the boundaries of what's possible with artificial intelligence.

Breakthrough Moments
Jan 25, 2024
15 min read
The Moment Everything Clicked: My First Neural Network Breakthrough
That 3 AM moment when my custom CUDA kernel finally worked—reducing training time by 70%. This is the story of persistence, failure, and the pure euphoria of a breakthrough that changed everything I thought I knew about AI optimization.
CUDANeural NetworksOptimization+1
By Unknown
Technical Adventures
Jan 22, 2024
18 min read
Dancing with Unsloth: My Love Affair with LLM Optimization
Why I fell in love with Unsloth and how it revolutionized my approach to fine-tuning. A deep dive into the elegant mathematics behind memory reduction and the art of making large models dance on limited hardware.
UnslothLLMFine-tuning+2
By Unknown
Origin Story
Jan 20, 2024
12 min read
My Journey: From Curiosity to AI Engineering Mastery
The personal story behind my passion for AI - from late nights debugging CUDA kernels to breakthrough moments with Unsloth optimization. This is my journey of pushing the boundaries of what's possible in AI.
Personal JourneyAI AdventureGrowth+1
By Unknown
Learning Moments
Jan 18, 2024
10 min read
When My RL Agent Outsmarted Me: A Humbling Experience
The day my reinforcement learning agent found a solution I never considered—and taught me that sometimes the student becomes the teacher. A reflection on emergent behavior and the beauty of letting AI surprise us.
Reinforcement LearningEmergent BehaviorHumility+1
By Unknown
Code Artistry
Jan 15, 2024
14 min read
CUDA Poetry: Writing Beautiful Parallel Code
There's an art to CUDA programming that goes beyond optimization. This is my exploration of writing GPU code that's not just fast, but elegant—where performance meets poetry in parallel processing.
CUDAParallel ComputingCode Beauty+1
By Unknown
Optimization Tales
Jan 12, 2024
12 min read
The TensorRT Revelation: Speed vs. Simplicity
My journey from PyTorch prototypes to lightning-fast TensorRT inference. How I learned to balance the raw speed of optimized engines with the elegance of research code—and why both matter.
TensorRTPyTorchInference+1
By Unknown

Let's Work Together

Interested in collaborating on AI projects or discussing cutting-edge technology ?
I'd love to hear from you.

Get In Touch

Whether you're looking to optimize AI models, implement cutting-edge solutions, or explore research opportunities, I'm always excited to discuss new challenges and innovations in AI engineering.

Email
[email protected]
LinkedIn
Connect with me
GitHub
View my code
Location
Nairobi, Kenya

Areas of Expertise

AI OptimizationCUDA ProgrammingModel DeploymentReinforcement LearningResearch
Send a Message
Let's discuss your AI project requirements