nickgeorge.net home on nick

Simple and Efficient SEC Filing Parsing Using Go

Tue, 30 Dec 2025 00:00:00 +0000

There are many SEC filings parsers, this one is mine.

Public companies in the US are required to file standardized disclosures with the SEC. Tracking those filings and extracting structured data from them is extremely useful, but surprisingly difficult to do. I run an application focused on biotech company activity, and SEC filings are a core input that can tell you a lot about what companies are up to. For example, the cover image above is a screenshot of an interactive plot of insider trades based on Form 4 from Wave Life Sciences.

Debugging 'Malformed Database Schema' in SQLite

Wed, 12 Nov 2025 00:00:00 +0000

I run a biotech investment and competitive intelligence tracking application. The app is a go app run on a bare-bones VPS with a SQLite database as the data store, dead simple.

I was recently working on a major rebuild of the database on a separate server (build-server). When I was satisfied, with the changes, I tried to deploy:

# on the build-server

sqlite3 rxdatalab.db "PRAGMA quick_check;"
# quick_check
# -----------
# ok
sqlite3 rxdatalab.db ".tables"
# list of tables...
# copy to s3:

aws s3 cp rxdatalab.db s3://bucket/rxdatalab.db.new
# OK

Then, copy to my prod server, and start the app:

Paddle Overlay Checkout

Sat, 05 Jul 2025 00:00:00 +0000

First, let’s make a frontend key and save it in our hugo.toml under a key like paddleSandboxClientSideJsToken (you’ll also have a paddleProdClientSideJsToken). Then, on your layouts/products/single.html template make a javascript snippet like so:

<script src="https://cdn.paddle.com/paddle/v2/paddle.js"></script>
<script type="text/javascript">
{{- $paddleToken := "" -}}
{{- $priceIds := slice -}}
{{ if hugo.IsServer -}}
Paddle.Environment.set("sandbox");
{{- $paddleToken = .Site.Params.paddleSandboxClientSideJsToke -}}
{{- $priceIds = .Params.product.testPaddlePriceIds -}}
{{- else -}}
{{- $paddleToken = .Site.Params.paddleProdClientSideJsToken -}}
{{- $priceIds = .Params.product.prodPaddlePriceIds -}}
{{- end }}
Paddle.Initialize({
token: '{{ $paddleToken }}',
checkout: {
settings: {
variant: 'one-page' // or multi-page
}
}
});
let itemList = [
{{ range $index, $id := $priceIds -}}
{ priceId: "{{ $id }}", quantity: 1 }{{ if ne (add $index 1) (len $priceIds) }},{{ end }}
{{ end }}
]
function openCheckout(items){
Paddle.Checkout.open({
items: items
});
}
</script>

I use hugo.IsServer to determine where I’m running and switch between my sandbox environment with the test variables and my production version with the live tokens/config.

Listening to Paddle Webhooks

Sun, 08 Jun 2025 00:00:00 +0000

First, create a sandbox account at https://sandbox-login.paddle.com/signup and a live account at https://login.paddle.com/signup. Yes, there are two separate accounts for testing and live. Use the same email address and business details.

Since Paddle is a merchant of record, they are more careful about who they approve than payment providers like stripe. It will take time and a few manual verification steps before they activate your main account, and you will need things like terms of service + refund policy, a privacy policy and a pricing/products page. To follow along here, you will be working with the test account.

Tracking Salary Transparency in Biotech Jobs with PharmaPayWatch.com

Tue, 14 Jan 2025 00:00:00 +0000

What can we learn from company job postings?

I’ve been collecting job posts from 13 major pharma and biotech companies since November 2024. At first, I was interested in providing accurate and timely information about new jobs to the community since LinkedIn is a dumpster fire.

However, it turns out there may be a lot of information you can learn about companies and the industry in general by watching how they hire!

NIH Should Buy 23andMe

Sun, 24 Nov 2024 00:00:00 +0000

23andMe is a remarkable company. Founded in 2006, they pioneered direct-to-consumer genetic testing and have since genotyped DNA from over 15 million individuals, making it the “world’s largest crowdsourced platform for genetic research”¹.

Importantly, over 80% of 23andMe customers have consented to the use of their data for research, creating a unique and invaluable resource for genetic and health studies.

The Unmatched Scale of 23andMe’s Data

The size of 23andMe’s collection is hard to appreciate without looking at other large initiatives such as NIH’s All of Us program, or the UK BioBank. 23andMe’s investor relations documents highlight the fact that its collection of genotyped individuals is several orders of magnitude larger than its closest competitor.

Static Dashboards

Mon, 04 Nov 2024 17:39:33 -0500

I’ve been building dashboards (essentially just pretty tables and some plots) that display data from different public data sources in a novel ways.

Most of these dashboards are ephemeral and experimental, so I’d rather not host a separate public application + database before I’m sure they’ll be around for a while. The simplest and most secure way to do this is to setup a static website and setup some kind of auto-update mechanism to build and re-deploy¹.

Open Knowledge - An Open Alex Explorer

Sun, 03 Nov 2024 00:00:00 +0000

OpenKnowledge

OpenAlex API Explorer

Enter search terms[info]:

Related terms:

Click a concept to search

man(ual) pages are all you need

Sun, 14 Apr 2024 00:00:00 +0000

With the ever decreasing quality of search thanks partly to ad-tech, google, and SEO (and more recently AI/LLM-enhanced SEO¹) I’m noticing that it is becoming much more difficult to find quality technical information online.

In the past, I defaulted to tutorial style learning, involving web searches for tutorials and “getting started with …” style articles, skim those and work from there.

This style was easy, beginner friendly, and great for starting out. Unfortunately, it has also become much more difficult if your topic is popular/mainstream (e.g., Python/JavaScript/LLM). You have to know a decent amount about the topic to even know if you are reading time wasting SEO garbage or useful information. If your topic of interest is more esoteric or niche and doesn’t have a lot of high value SEO keywords you might get lucky and find quality information– but there is also a better way!

Bootstrapping Ansible on FreeBSD

Tue, 02 Apr 2024 00:00:00 +0000

Automating tasks with Ansible has been a great way to learn about server management best practices and get Linux servers up and running quickly. When I started experimenting with FreeBSD¹ in my homelab I was surprised that Ansible wasn’t well supported.

Python and sudo

The primary difficulty is that while Ansible is “agentless”, it does require Python on the target server. FreeBSD doesn’t clutter up your system with interpreters you may not need², so using Ansible won’t work out of the box. After some fiddling, here is my basic ‘bootstrap’ playbook for FreeBSD. Note that FreeBSD also does not have sudo by default, so I also install that and initially use su to root for privilege escalation³ when bootstrapping.