Publication
Engineering journal for builders.
I write about things I ship and learn: systems design, AI implementation, tutorials, and practical engineering decisions.
- Articles
- 15
- Reading
- 61 min
- Updated
- August 15, 2026
Articles
Filter by domain, or start with the latest.
Reading Dense Python: Comprehensions, zip(), enumerate(), and the * Operator
One line of Python can stack five separate ideas on top of each other. Here is how to take that line apart from the inside out, so the shorthand stops looking like magic and starts looking like composition.
- August 7, 2026AI & Systems Architecture
Designing an Air-Gapped RAG Architecture for Healthcare
A hospital wants a GPT-4-class assistant, but clinical data can never leave infrastructure it controls. That single constraint disqualifies every hosted API and forces a specific architecture. Here is that architecture, box by box.
10 min readDeep DiveAdvanced - August 6, 2026Engineering Systems
From a Home Router to a Hospital Network: Understanding Enterprise Networking Step by Step
A switch connects devices. A router connects networks. A firewall decides which connections are allowed. Here is how those four ideas scale from the box on your desk to a hospital with its own datacenter.
9 min readDeep DiveIntermediate - August 1, 2026Engineering Systems
Demystifying Certbot, SSL, and NGINX: A Complete Guide for Full-Stack Developers
Your senior developer tells you to just run Certbot, a few files appear, and the padlock shows up. Here is what actually happens between Namecheap, your server, Let's Encrypt, and NGINX.
3 min readTutorialIntermediate - July 31, 2026AI & Systems Architecture
IVFFlat vs HNSW in pgvector: The Difference That Finally Made It Click for Me
IVFFlat partitions the vector space; HNSW navigates a layered graph. Getting the two mental models confused is how a production query ended up at ef_search = 600 and a 4,900 ms latency cliff.
5 min readDeep DiveAdvanced - July 18, 2026AI & LLM
Traditional RAG vs Document Intelligence: The Next Evolution of Enterprise AI
Most AI document products stop at upload-ask-answer. Operational workflows need more than answers — they need complete, structured, auditable data, with RAG as one component inside a larger system rather than the whole product.
2 min readDeep DiveAdvanced - July 11, 2026AI & LLM
DETR vs. ViT: Two Different Questions, Not One Architecture Plus a Head
A Vision Transformer classifies an image. DETR finds every object in it. The two models diverge earlier than most explanations admit, and the distinction changes how you should read both papers.
3 min readDeep DiveIntermediate - July 2, 2026AI & LLM
RAG Does Not Do Math, and That's Okay
Retrieval gets your model the right numbers. It does not make the model good at arithmetic. Here is how to stop pretending otherwise.
3 min readField NotesIntermediate - June 30, 2026AI & Systems Architecture
Not Every Workflow Needs an AI Agent: Choosing the Right Autonomy Level
Every startup wants autonomous AI agents that think and act independently. But unconstrained autonomy introduces risk. Here is how to select the right operational autonomy level for your product.
2 min readFrameworkIntermediate - June 29, 2026AI & Systems Architecture
Choosing Your Automation Boundary: Tasks, Processes, or Systems?
Automatically extracting data from an invoice is vastly different from automating an enterprise finance operation, yet both are loosely labeled "AI automation." Here is how to define clear technical scope boundaries before writing code.
2 min readFrameworkIntermediate - June 18, 2026AI & LLM
Fine-Tuning Is Fine-Tuning Until It Isn't
Fine-tuning looks like a solved problem right up until the model quietly forgets something you never tested for. A field guide to when it helps and when it bites back.
3 min readField NotesIntermediate - June 4, 2026Field Notes
The 5 Questions I Ask Before I Trust Any AI System
A short, repeatable framework for deciding whether an AI feature is ready for production, before a user finds the gap for you.
3 min readFrameworkIntermediate - May 21, 2026AI & LLM
The Single Pipeline That Explains Modern Computer Vision
Classification, detection, segmentation, and vision-language models all reduce to the same five stages. Once you see the pipeline, the papers stop looking like separate fields.
3 min readDeep DiveIntermediate - May 7, 2026Engineering Systems
How We Built Map-Based Property Search in East Africa Without Google Maps
Google Maps pricing and coverage gaps make it a bad fit for proptech in East Africa. Here is the OpenStreetMap-based stack we used instead, and what broke along the way.
3 min readCase StudyAdvanced - April 7, 2024Engineering Systems
The Power of Static Typing in Production Systems
A practical guide to how static typing improves reliability, refactoring safety, and delivery speed in real production codebases.
2 min readTutorialIntermediate
Need an article on a specific topic?
I can publish breakdowns on requested domains: backend, AI systems, frontend architecture, or short tutorials.