If you enjoy my content, it would mean the world to me if you invited friends to subscribe and read with us. If you refer friends, you will receive benefits that give you special access to my expertise and community.

Subscribe now

How to participate

1. Share American engineered intelligence. When you use the referral link below, or the “Share” button on any post, you'll get credit for any new subscribers. Simply send the link in a text, email, or share it on social media with friends.

Refer a friend

2. Earn benefits. When more friends use your referral link to subscribe, you’ll receive special benefits.

• Get Access to subscriber chat for 3 referrals

• Get Access to my private slack for 5 referrals

• Get Founding membership for 25 referrals

Visit the leaderboard

To learn more about “refer a friend,” check out Substack’s FAQ.

I hope to spread accurate knowledge about Artificial Intelligence and also to connect with businesses who are seeking to implement it. Thank you for helping get the word out about American engineered intelligence!

This is a concierge model training service. I run your workloads myself on private, high-performance infrastructure that I control. If needed, I can supplement that hardware with cloud resources or additional leased machines to match your project requirements.

Whether you're fine-tuning a language model, experimenting with diffusion, or running a new reward model from scratch, I help you get results fast without the typical overhead.

What I Offer

This is a full-service training solution. You do not need to bring code, data, or a model architecture. Just describe what you want the model to do, and I will take it from there.

I can design the system, implement the training pipeline, clean or generate data, run the training process, and deliver clean, reproducible artifacts ready for deployment.

Ideal clients include:

• Startups and founders who want to integrate AI into a product

• Researchers exploring novel model behaviors or domains

• Independent developers who have a goal in mind but need help getting there

My Infrastructure

Primary compute runs on my personal training cluster:

• 2x NVIDIA RTX 5080 GPUs

• AMD Ryzen 9950X CPU

• 192 GB DDR5 RAM

• 2 TB NVMe SSDs and 8 TB of cold storage

• 2.5 Gbps ethernet for fast artifact transfer

I also have the ability to orchestrate additional GPU nodes in the cloud or through specialized rental providers. If your job needs more hardware, I will handle acquisition, setup, and orchestration myself. You will not need to worry about the infrastructure.

Supported Workloads

I support:

• Language model fine-tuning and embedding models

• Diffusion models for text-to-image, audio, or video

• Reward modeling, policy tuning, and RLHF

• Custom architectures built in PyTorch or JAX

If your model can fit within a multi-GPU setup, I can train it, optimize it, and package it for use.

My Approach

You will work directly with me. I run the training job, monitor it, adjust as needed, and deliver results. I can use your preferred tooling or suggest the best approach for your use case.

Everything runs in a controlled environment. You get clear documentation, reproducible scripts, and ready-to-integrate output.

If you are not sure where to begin, I will help you define the objective and scope the project. This is a partnership, not a one-size-fits-all platform.

Pricing and Guarantee

I charge $300 per hour of my time. That includes all compute, networking, and power costs for jobs that I can run on my own hardware or in your cloud environment.

I stand behind my work with My Iron-clad guarantee. If you are not satisfied with the results, you do not pay.

Work With Me

I work primarily with clients in the United States.

You get my attention, my compute, and my expertise focused entirely on your problem. No queue. No subcontractors. No wasted cycles.

Book a call today and scope your training needs.

I help companies design and operate modern, secure, high-performance systems in the cloud. Whether you're shipping a consumer app or deploying AI in production, I give you the foundations to move fast, scale safely, and stay in control.

What I Offer

I provide deep expertise across both cloud-native architecture and modern MLOps. That means I can design the infrastructure your applications rely on and build the automated pipelines that keep your machine learning systems healthy and efficient.

Cloud-Native Application Architecture

I build scalable, cost-effective infrastructure on AWS and Azure. Whether you're starting greenfield or modernizing legacy systems, I deploy everything using Infrastructure as Code and clean CI/CD practices.

Key capabilities:

• Cloud cost optimization and billing audits

• Greenfield infrastructure design and bootstrapping

• High-availability full-stack architectures

• Deployment of serverless systems, container clusters, and AI workloads

• AWS Well-Architected Reviews

MLOps, DevOps, and DevSecOps

Once your model is trained, the real work begins.

I build robust pipelines for training, evaluation, deployment, and monitoring—backed by best practices from enterprise ML systems.

Key capabilities:

• Secure, automated pipelines for training and deployment

• Optimization of training jobs for cost, throughput, and accuracy

• Inference endpoint deployment and autoscaling

• Integrated logging, metrics, and tracing for full observability

• Software supply chain security and artifact hygiene

My Approach

I build systems that are clean, observable, and designed for scale.

You get reproducible infrastructure, secure defaults, and zero guesswork when things break. My philosophy is to move fast—but never recklessly.

What I prioritize:

• Automation first — deployment through CI/CD, not manual clicks

• Security by default — proper IAM, secrets management, and audit trails

• Performance and cost — tune where it matters, simplify where it doesn’t

• Observability — full visibility from API to database to GPU

Results You Can Expect

Clients typically see:

• 30 to 70 percent reduction in cloud spend

• 3x faster deployment cycles after adopting CI/CD

• Lower time to resolution and higher uptime through better observability

• Training costs cut in half with smarter infrastructure and job scheduling

Case Study

Event-Driven Invoice Processing at Scale

A global enterprise needed to process over 86 million invoice-related events per day with strict SLAs, audit requirements, and near real-time visibility.

I led the architecture and implementation of a fully serverless Business Event Monitoring System (BEMS) on AWS.

The solution uses API Gateway, EventBridge, SNS, SQS, and Lambda to process events in partitioned pipelines across regional deployment cells. Events are streamed into Amazon Timestream and visualized in QuickSight dashboards that power executive and operational insights.

The system includes dead-letter queues, API usage plans, anomaly detection, and full end-to-end observability with CloudWatch and X-Ray. It operates with zero idle compute, handles regional failover, and scales to over 600,000 events per minute.

Result: Real-time visibility into invoice flow, 80 percent faster recovery from stuck events, and zero operational intervention during end-of-month spikes.
Client feedback: “We can finally see what’s happening and trust that it’s right.”

Pricing and Guarantee

I charge $300 per hour of my time. That includes all compute, networking, and power costs for jobs that I can run on my own hardware or in your cloud environment.

I stand behind my work with My Iron-clad guarantee. If you are not satisfied with the results, you do not pay.

Work With Me

I work primarily with clients in the United States.

You get my attention, my compute, and my expertise focused entirely on your problem. No queue. No subcontractors. No wasted cycles.

Book a call today and scope your training needs.

You need systems aligned with your workflows, your goals, and your strategy.

That’s where I can help.

What I Offer

I specialize in custom AI integration. From prototype to production deployment.

Whether you are using LLMs, diffusion models, or classical ML, I will help you architect, optimize, and ship high-impact systems that run fast and scale cleanly.

End-to-End Integration

• Model selection and fine-tuning

• Deployment to cloud, on-prem, local or embedded hardware

• Seamless integration with APIs, pipelines, and your existing stack

Supported Workloads

I support a wide range of AI workloads tailored to your needs. This includes large language models like the GPT family, Mistral, LLaMA, Claude, and other open-weight alternatives.

I also work with diffusion models for image, video, and audio generation. For edge deployments, I ship quantized models that run efficiently on devices like iPhones, Jetsons, Raspberry Pi, and other constrained hardware environments.

Use Cases

• Automating route optimization in logistics

• Deploying LLM copilots that cut internal workload in half

• Personalizing on-device user experiences with local inference

My Approach

I bring an infrastructure-first mindset. I design for performance, security, and scale from day one. You get robust systems, not just model code.

• Modern infrastructure using Infrastructure as Code, Containers, and CI/CD

• Security and observability integrated from the start in end-to-end workflow

• Optimized performance with low-latency inference and cost-efficient runtime

Results You Can Expect

My clients see:

• Lower infrastructure costs

• Shorter development cycles

• More responsive, intelligent software

All backed with a 100% satisfaction guarantee.

Example Wins

Here are some wins from my career so far:

• Inference latency (p90) reduced by 70% with autoscaling

• $50,000 per month saved after token spend optimization

• 2x conversion on apps personalized with lightweight local models

Let’s create an even bigger impact with your use case.

Work Sample

Local AI Copilot for Musicians

Musicians love to use offline, low-latency, privacy-respecting software in the studio. musegpt allow users to chat with AI while they make music in popular production software like Ableton, Pro Tools, or Logic. It deploys an LLM locally inside a llama.cpp webserver exposed through JUCE’s VST plugin interface.

Result: Fully offline AI chat experience inside existing music software
User feedback: “This is the first AI tool that actually helps in the studio”

Pricing and Guarantee

I charge $300 per hour of my time. That includes all compute, networking, and power costs for jobs that I can run on my own hardware or in your cloud environment.

I stand behind my work with My Iron-clad guarantee. If you are not satisfied with the results, you do not pay.

Work With Me

I work primarily with clients in the United States.

When you hire me, you get direct access to my expertise in systems, models, and production deployment. No subcontractors. No layers.

What are you waiting for? Book your free intro call today.

Subscribe now

Starting over is hard.

In the past few years, I’ve been blessed with many opportunities that let me travel all across the United States. From the gentle fog rolling down San Francisco Bay, to the gleaming surface of the Puget Sound, with brief stops in Vegas, Utah, and Denver.

The pace of innovation is rapid, and frankly, life has been hectic. Between multiple stints at AWS, cofounding multiple startups, and consulting on the road, we stay busy.

I’m founding agentlab.us to organize and coordinate my business activity as an independent contractor, but more on that later. Let’s talk about why.

What makes today special in 2025

I believe Artificial Intelligence (and more broadly, Machine Learning) technology provides the radical capability to reshape the American way of life in a positive way.

I see these tools being used to increase agency, individuality, and economic participation in ways we can’t imagine today. As an engineer, scientist, and technologist, I will further these technologies to benefit the American public.

I chose today, Memorial Day, to create my firm for a reason. As we reflect on our past as a nation and honor our fallen heroes, I found myself considering how best to honor that sacrifice. I can think of no better way than exercising my economic freedoms.

A community of American innovators

Let this be a space for a thriving community of American innovators from all walks of life. I aim to share thoughtful, direct, and honest case studies and think-pieces on AI progress and innovation. I will also provide open source code and open datasets for the public good. My intention is to grow my community, to support the open sharing of knowledge, and to ultimately be of use to the American public.

What you can expect

• Occasional public posts will include open source tools, public datasets, how-to guides, and other resources in the public interest.

• Subscribers will enjoy quarterly case studies including in-depth analysis of my work in artificial intelligence, customer benefits, and return on investment.

• I will not be accepting paid subscriptions at this time. It is my hope that sharing my knowledge and work publicly might benefit others while also attracting interesting and challenging short-term gigs my way.

Expressing my gratitude

Thank you, dear reader. For all that you do. Stay true.

Reach out if you need me via email or the button above.