Python

Last quarter I had to validate a nightly export pipeline that dropped 120 CSVs into a staging folder before a billing cutover. A quick diff of folders wasn’t enough; I needed to prove that specific files matched across two locations and surface the handful that didn’t. That’s exactly where Python’s filecmp.cmpfiles() shines. It compares a […]

Last month I was debugging a vibration sensor pipeline for a manufacturing line. The device logged at irregular intervals, while our model expected a clean 100 ms grid. I could have pulled in a large interpolation package, but I only needed straight lines between known points. In moments like this, I reach for numpy.interp, a

Last year I was troubleshooting a data pipeline on a CI runner. Locally it worked, but on the runner it silently used /usr/bin/python3.8 while my laptop used a venv on Python 3.12. The script parsed dates differently and a small bug slipped in. The fix was simple: log the full interpreter path and fail fast

Last month I was reviewing sign-up sources for a small SaaS launch. The export had hundreds of rows, but the question from my PM was simple: which three channels delivered most of the sign-ups, and what share did each take? For a quick briefing, a pie chart still works well. A pie chart is a

Last quarter I had a pipeline that looked fine until a CFO asked why the median salary was 0. The CSV had salaries like ‘$95,000‘, pandas imported them as object strings, and my math silently concatenated text. That kind of quiet error is why I treat dtype work as first‑class. When you convert a Series

Last week I was cleaning up a deployment script that decides whether a feature flag should be enabled for a release candidate. The logic was correct, but it was scattered across three tiny if blocks that each assigned a value and moved on. In places like that I want the value, not the ceremony. Python‘s

Last year I had to build a small internal dashboard that tracked which tutorial videos our team had referenced in code reviews. The videos were scattered across playlists, and everyone described them differently. What I really needed was a reliable, programmatic way to ask a video, ‘Who uploaded you? How long are you? What’s your

Most of my Colab sessions start the same way: a notebook with a great idea and an empty filesystem. The dataset I want is on Kaggle, but Colab resets and forgets everything. I used to spend 10–15 minutes per session downloading, unzipping, and hunting for the file path. That’s wasted time and a recipe for

Last quarter I was debugging a feature store pipeline that scored incoming events. The logs showed arrays of weights, but every run looked different because the print settings were inherited from whatever notebook or service ran before. I needed a dependable string snapshot of the data without a lot of extra metadata, and I needed

When I teach real-time programming, I often start with Pong because it is a tiny problem that reveals every big idea: input, physics, timing, and scoring. Pong began in the early 1970s as a table-tennis themed arcade game with two paddles and a ball, and that simplicity still makes it a perfect lab. You can

Last year I was on call for a rollout that randomly exposed a new checkout flow to 1% of traffic. The logic was just “pick a user ID from a list,” but it sat in a high‑impact path, so every detail mattered. The way you choose a random element affects reproducibility, fairness, and how you

Last quarter I inherited a Python service that ran fine on my laptop but failed in CI because another developer had installed a newer transitive library. The bug was not in our code; it was in our assumptions. That kind of drift is why I treat dependency management as a core engineering task, not a