Creating grounded AI experts
Creating grounded AI experts
Date: 2025-09-30
Hello!
I'm Mark — a machine learning engineer working at the intersection of AI, automation, and online systems. At Northwestern, I lead ML infrastructure for academic research on social media, studying how people communicate and influence each other online. Outside academia, I consult with organizations on anything AI: building LLM agents, automating workflows, designing custom recommendation engines, and making data work end to end. I bridge research and engineering to help teams ship smart, scalable systems.
Recent posts
The hard work comes before you start asking ChatGPT to write (code, papers, docs, etc.)
The hard work comes before you start asking ChatGPT to write (code, papers, docs, etc.)
Date: 2025-09-29
Quick tips that help me work better with LLMs
Quick tips that help me work better with LLMs
Date: 2025-09-28
What I’m reading today (2025-09-26)
What I’m reading today (2025-09-26)
Date: 2025-09-25
Using AI to create simple demos and to get unstuck quickly
How I use AI to create simple demos and get unstuck quickly
Date: 2025-09-22
My MCP Stack: The Tools That Currently Power My AI Workflow
My current MCP Stack: The Tools That Power My AI Workflow
Date: 2025-09-18
AI Agent Workflows as GPU Architecture: A Parallel Computing Analogy
AI Agent Workflows as GPU Architecture: A Parallel Computing Analogy
Date: 2025-09-17
From Production Crashes to Systematic Solutions: How AI-Human Collaboration Solves Complex Problems
From Production Crashes to Systematic Solutions: How AI-Human Collaboration Solves Complex Problems
Date: 2025-06-29
The surprising effectiveness of SQLite for non-persistent use cases.
I’ve been using SQLite as a buffer for some microservices that I’m building out. Normally I would’ve thought of using something like SQS or Kafka for somethi...
Date: 2025-01-30
If you’re going to do something, you might as well do it well
“If you’re going to do something, you might as well do it well”
Date: 2025-01-21
Why do we teach developers that serverless is always better?
Why do we teach developers that serverless is always better?
Date: 2025-01-21
Setting up zsh on an HPC environment
Setting up zsh on an HPC environment
Date: 2025-01-14
Knowing how to code isn’t enough to keep you employed in tech
Knowing how to code isn’t enough to keep you employed in tech The promise of just doing a CS degree or a bootcamp and walking into a six-figure tech job, whi...
Date: 2024-06-23
You don’t have to be the best at what you do. You just have to be good at fixing people’s problems.
You don’t have to be the best at what you do. You just have to be good at fixing people’s problems. I am often too hard on myself. I see all the ways that I ...
Date: 2024-06-11
Money won’t open your textbooks or lift your weights.
Money won’t open your textbooks or lift your weights.
Date: 2024-06-02
Research updates, 2024-06-01
Research updates, 2024-06-01
Date: 2024-06-01
Research updates, 2024-05-31
Research updates, 2024-05-31
Date: 2024-05-31
Research updates, 2024-05-30
Research updates, 2024-05-30
Date: 2024-05-30
Accidentally deleting prod data (and the importance of backups)
Accidentally deleting prod data (and the importance of backups).
Date: 2024-05-28
Experiments with LLM classification for political content (Part IX)
Using LLMs to classify if social media posts are political or not (Part IX). I’m working on a project that involves gathering social media posts from Bluesky...
Date: 2024-05-21
Experiments with LLM classification for political content (Part VIII)
Using LLMs to classify if social media posts are political or not (Part VIII). I’m working on a project that involves gathering social media posts from Blues...
Date: 2024-05-21
What I’m learning (2024-05-20)
What I’m Learning (2024-05-20 edition)
Date: 2024-05-20
Experiments with LLM classification for political content (Part VII)
Using LLMs to classify if social media posts are political or not (Part VII). I’m working on a project that involves gathering social media posts from Bluesk...
Date: 2024-05-18
Brief thoughts on outsourced labor
I saw a Twitter post about someone explaining how they found out “this one weird trick” in entrepreneurship, where they discovered they could hire Filipinos ...
Date: 2024-05-17
Why I’m not too worried about AI taking over the world
Why I’m not too worried about AI taking over the world [stuff stuff]
Date: 2024-05-17
Experimenting with DSPy, Pt. I
Experimenting with DSPy, Pt. I As part of me learning more about LLMs, I’m working through different use cases, tools, and examples that I find online. Right...
Date: 2024-05-16
Experimenting with LlamaIndex, Pt. I
Experimenting with LlamaIndex, Pt. I As part of me learning more about LLMs, I’m working through different use cases, tools, and examples that I find online....
Date: 2024-05-16
LLMOps self-education, Pt. III
Session III: Creating a more robust RAQA system using LlamaIndex I’m working on improving my RAG-building skills. As a part of that, I’m taking courses onlin...
Date: 2024-05-15
Riffs about higher education
Riffs about the value of higher education I’m lucky enough to have gone to college at a Ivy League school on a full scholarship. It definitely changed my lif...
Date: 2024-05-15
LLMOps self-education, Pt. II
Session II: Questioning Barbie and Oppenheimer Through the Use of Agents I’m working on improving my RAG-building skills. As a part of that, I’m taking cours...
Date: 2024-05-14
LLMOps self-education, Pt. I
Session I: Course Intro | Building Your First RAG App with LangChain, Chainlit, and Hugging Face I’m working on improving my RAG-building skills. As a part o...
Date: 2024-05-14
What I’m learning (2024-05-14)
What I’m Learning (2024-05-14 edition)
Date: 2024-05-14
What I’m learning (2024-05-09)
What I’m Learning (2024-05-09 edition)
Date: 2024-05-09
Experiments with LLM classification for political content (Part VI)
Using LLMs to classify if social media posts are political or not (Part VI).
Date: 2024-05-08
Experiments with LLM classification for political content (Part V)
Using LLMs to classify if social media posts are political or not (Part V).
Date: 2024-05-08
Experiments with LLM classification for political content (Part IV)
Using LLMs to classify if social media posts are political or not (Part IV).
Date: 2024-05-03
Experiments with LLM classification for political content (Part III)
Using LLMs to classify if social media posts are political or not (Part III).
Date: 2024-05-02
Experiments with LLM classification for political content (Part II)
Using LLMs to classify if social media posts are political or not (Part II).
Date: 2024-05-01
Experiments with LLM classification for political content (Part I)
Using LLMs to classify if social media posts are political or not (Part I).
Date: 2024-04-30
Summary of “Algorithm-mediated social learning in online social networks”
Paper overview
Date: 2023-12-08
Working at a brick-and-mortar business as a tech bro
In my ~2 month stay in America, I’ve stumbled upon a volunteer unpaid job as a PM (I guess?) helping my mother and brother run two restaurants.
Date: 2023-12-04
Creating custom feeds in Bluesky: a data scientist’s experience
Currently, as part of my research, I am investigating the possibility of manipulating people’s social media algorithms and measuring their perceptions of pol...
Date: 2023-11-29
Working with Bluesky as a researcher
I’m currently doing research at Northwestern, studying how people engage with political content online on social media, especially how they relate to algorit...
Date: 2023-11-28
‘Hello World!’ (properly this time)
This is a proper Hello World if I’ve ever seen one.
Date: 2023-10-03