AI / ML / Automation / Civil Engineering
Deploying AI
Some notes on AI deployment from Sol Rashidi’s book ‘Your AI Survival Guide: Scraped Knees, Bruised Elbows, and Lessons Learned from Real-World AI Deployments‘. She’s probably one of the best AI deployment practitioners out there. She’s a C-Suite executive (Chief Data/AI/Analytics Officer) at several Fortune 100’s, has 10 patents, and
SQL Dump
Crib notes from when I used SQL to manage my online platform’s database. Focuses on the most practical 20% that delivers 80% of the results.
Automating Construction News
Reading industry news is part of the job. But doing it manually every day—clicking headlines, skimming paragraphs, filtering out noise—is a time sink. So I automated it. The script checks sites like ICE, New Civil Engineer, Construction Enquirer etc., pulls the latest articles, summarises them with GPT, and compiles it all
Unsupervised Machine Learning (TBM)
I’ve mostly toyed with supervised machine learning in the past due to its much larger utility compared to unsupervised learning. But, this has left me knowing very little about unsupervised machine learning. Although it has lower utility, I still want to know how it can be used practically, especially for
April 2023: Prompt Engineering Notes
Guessing this will quickly get outdated with how fast the LLM space is moving, but my notes on prompt engineering below. Uploading as a quick reference source as otherwise it will 100% get lost on my desktop!
Making a House Screening Tool
When I started looking at houses, one thing was clear: location matters. A lot! But it’s hard to get a feel for a place without actually living there. Crime stats seemed like a good proxy for quickly finding out what the quality of life is like in an area without
Training a YOLOv8 Object Detector
If you want to teach a model to spot things in pictures, then object detection is probably what you’re looking for. Computer vision isn’t exactly anything new, but YOLOv8 is leaps and bounds better than earlier versions I used in the past, which makes this interesting. To help explain the
Embeddings
My note dump on OpenAI’s embedding model. First impressions – it looks really promising and is much faster than using local embeddings from HuggingFace models. Will be using it for vector similarity matching to auto-group KWs for my online platform’s SEO articles.
Automating Document Comparison w/ Python & Tkinter
So, you’ve got a client with some seriously specific document requirements, huh? I feel your pain! Processes are important, especially for change tracking, but sometimes they can feel like they’re eating up all your time! I recently worked on a project where the client needed the documents to be updated
Earthwork notes from SHW & Nowak/Gilbert
My WIP notes from: Not a full summary at all, just a collection of the key points I’ve found useful for working on earthwork projects that deal with things like compaction, classification, fill behaviour etc. etc. You’ll probably need to zoom in in order to be able read anything.
Automating Passport Appointment Checks
My partner recently needed to visit a very unwell family member overseas. But her current passport was about to expire, and we needed to renew it ASAP. The UK Government website has a way to fast-track the process, but it’s near-impossible to actually secure a slot. The slots appear randomly
Categorising Data with Zero-Shot Classification
Recently, I needed to group about 1000 terms into a number of categories. Doing it manually was far too painful, so I turned to machine learning as it can be so good for these kinds of tasks. Specifically, I used the torch and transformers library from Hugging Face. Their pipeline
March 2021: ML Notes
Dump of my notes from Machine Learning for Absolute Beginners (3rd e.d.) by Oliver Theobald. V. good book for a quick intro to ML.
Fixing Broken SharePoint Links
If you’ve ever worked with SharePoint, you’ve probably seen this happen: Someone moves a folder. Or renames a file. And now the link doesn’t work… Now you’re stuck manually going through each folder you think it might be in, hoping you get lucky or can suss out where it might
Reducing Analysis Time by 60% by hacking an obsecure LUSAS feature
In one of my recent bridge assessment projects, I ran into some serious inefficiency. As part of a Level 2 Finite Element Analysis (FEA), I needed to run over 10 different analysis scenarios. Each scenario included four load cases (LCs): The problem was that LC1 to LC3 were exactly the
Deploying AI
Some notes on AI deployment from Sol Rashidi’s book ‘Your AI Survival Guide: Scraped Knees, Bruised Elbows, and Lessons Learned from Real-World AI Deployments‘. She’s probably one of the best
SQL Dump
Crib notes from when I used SQL to manage my online platform’s database. Focuses on the most practical 20% that delivers 80% of the results.
Automating Construction News
Reading industry news is part of the job. But doing it manually every day—clicking headlines, skimming paragraphs, filtering out noise—is a time sink. So I automated it. The script checks sites
Unsupervised Machine Learning (TBM)
I’ve mostly toyed with supervised machine learning in the past due to its much larger utility compared to unsupervised learning. But, this has left me knowing very little about unsupervised
April 2023: Prompt Engineering Notes
Guessing this will quickly get outdated with how fast the LLM space is moving, but my notes on prompt engineering below. Uploading as a quick reference source as otherwise it
Making a House Screening Tool
When I started looking at houses, one thing was clear: location matters. A lot! But it’s hard to get a feel for a place without actually living there. Crime stats
Training a YOLOv8 Object Detector
If you want to teach a model to spot things in pictures, then object detection is probably what you’re looking for. Computer vision isn’t exactly anything new, but YOLOv8 is
Embeddings
My note dump on OpenAI’s embedding model. First impressions – it looks really promising and is much faster than using local embeddings from HuggingFace models. Will be using it for
Automating Document Comparison w/ Python & Tkinter
So, you’ve got a client with some seriously specific document requirements, huh? I feel your pain! Processes are important, especially for change tracking, but sometimes they can feel like they’re
Earthwork notes from SHW & Nowak/Gilbert
My WIP notes from: Not a full summary at all, just a collection of the key points I’ve found useful for working on earthwork projects that deal with things like
Automating Passport Appointment Checks
My partner recently needed to visit a very unwell family member overseas. But her current passport was about to expire, and we needed to renew it ASAP. The UK Government
Categorising Data with Zero-Shot Classification
Recently, I needed to group about 1000 terms into a number of categories. Doing it manually was far too painful, so I turned to machine learning as it can be
March 2021: ML Notes
Dump of my notes from Machine Learning for Absolute Beginners (3rd e.d.) by Oliver Theobald. V. good book for a quick intro to ML.
Fixing Broken SharePoint Links
If you’ve ever worked with SharePoint, you’ve probably seen this happen: Someone moves a folder. Or renames a file. And now the link doesn’t work… Now you’re stuck manually going
Reducing Analysis Time by 60% by hacking an obsecure LUSAS feature
In one of my recent bridge assessment projects, I ran into some serious inefficiency. As part of a Level 2 Finite Element Analysis (FEA), I needed to run over 10