<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Emily Ekdahl on Medium]]></title>
        <description><![CDATA[Stories by Emily Ekdahl on Medium]]></description>
        <link>https://medium.com/@emekdahl?source=rss-3b9a117391dd------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*IeQfAEOVKJUwyF8nfCq20A.png</url>
            <title>Stories by Emily Ekdahl on Medium</title>
            <link>https://medium.com/@emekdahl?source=rss-3b9a117391dd------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 09:24:12 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@emekdahl/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Why Your AI Music Prompts Aren’t Working (And What To Do Instead)]]></title>
            <link>https://emekdahl.medium.com/why-your-ai-music-prompts-arent-working-and-what-to-do-instead-2183e33c49c1?source=rss-3b9a117391dd------2</link>
            <guid isPermaLink="false">https://medium.com/p/2183e33c49c1</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[music]]></category>
            <category><![CDATA[writing-prompts]]></category>
            <category><![CDATA[creativity]]></category>
            <dc:creator><![CDATA[Emily Ekdahl]]></dc:creator>
            <pubDate>Tue, 02 Dec 2025 14:02:01 GMT</pubDate>
            <atom:updated>2025-12-05T00:41:10.626Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*UMfBNEQVQkhCzHqK" /><figcaption>Photo by <a href="https://unsplash.com/@siednji?utm_source=medium&amp;utm_medium=referral">Siednji Leon</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h3>High-Level Creation Flow</h3><blockquote>Idea → Messy Draft → Critique → Rewrite → Iterate in Suno → Remix → Edit → Refine → Publish</blockquote><h3>Model Specialization</h3><p><strong>Claude Opus: </strong>Best for emotional writing, humor, and natural line flow</p><p><strong>ChatGPT 5.1: </strong>Brainstorm, outline, generate Suno prompts</p><p><strong>Suno: </strong>Generate music and refine sound</p><p><strong>DALL-E: </strong>Generate album art and maintain a consistent visual style</p><h3>Set Up for Success</h3><h4>Context Engineering</h4><p><strong>Use Projects: </strong>Create a dedicated project in your chat AI to serve as a persistent context for your work.</p><p><strong>Define the Goal:</strong> Clearly explain your overall objective for the project in the description.</p><p><strong>Enrich the Context: </strong>Add relevant images or documents to help the AI understand your vision.</p><p><strong>Maintain History:</strong> Continuously update the working document (including all lyrics) in your project. This allows the chat AI to track high-level themes, avoid repetition, and maintain context without requiring constant copy-pasting.</p><h4>Prompt Engineering</h4><p><strong>Clarity and Concision:</strong> Be as clear and concise as possible when communicating your requests. While examples are often helpful, more detail is not always better.</p><p><strong>Iterative Feedback: </strong>Provide the AI with specific feedback on its ideas early and often to guide its output.</p><h3>Brainstorm</h3><h4>Album Concept</h4><p>Brainstorm with your chat AI to define the core of your album</p><p><strong>Theme/Idea: </strong>Define the high-level image, theme, or idea you want to convey.</p><p><strong>Genre: </strong>Identify the primary genre(s) you are drawn to.</p><p><strong>Genre Analysis: </strong>Research common themes and messages within that genre. Decide whether you will adhere to or parody those messages.</p><h4>Setlist Structure</h4><p><strong>Template Selection:</strong> Use a genre-specific template to outline your album setlist.</p><p><strong>Song Count: </strong>Determine the typical number of songs for that template.</p><p><strong>Emotional Arc: </strong>Outline the emotional flow of the album.</p><p><strong>Song Placement: </strong>Decide where to strategically place your strongest songs.</p><p><strong>Flow: </strong>Use one-line descriptions for each track slot to establish the album’s flow before writing full lyrics.</p><h4>Album Title</h4><p>Request ten title iterations from the AI, providing your setlist as context, and iterate until you are satisfied.</p><h3>Lyrics</h3><h4>Drafts and Structure</h4><p><strong>Initial Idea: </strong>Begin with the high-level concept for a specific song from your setlist.</p><p><strong>Narrative Arc: </strong>Brainstorm the song’s narrative flow with the AI.</p><p><strong>Lyrics Generation: </strong>Generate a messy draft</p><p><strong>Outline Lyrics for Suno:</strong> Have the AI outline the lyrics with structural markers (e.g., Verse 1, Pre-Chorus, Chorus, Bridge, Outro). This helps Suno do a better job of interpreting your lyrics.</p><h4>Refinement and Quality</h4><p><strong>Length: </strong>Prefer shorter songs. While v5 models can achieve 3—4 minutes, quality often degrades over extended runtimes.</p><p><strong>Cadence:</strong> Read your lyrics aloud before running them through Suno to check the natural rhythm and flow.</p><h4>Critique Lyrics with AI</h4><p>Ask Claude Opus to critique the lyrics, focusing on:</p><ul><li>Cohesive narrative arc</li><li>Excessive repetition (within or between songs).</li><li>Line and section length variation.</li><li>Overall song length.</li><li>Potential miscommunications (jokes that won’t land).</li><li>Vocalist challenges (difficult-to-pronounce or ambiguous words, excessive syllables).</li></ul><h4>Lessons Learned</h4><p><strong>Playtest Early: </strong>Run lyrics through Suno with minimal music prompting at least once. This quick run often reveals needed lyric revisions.</p><p><strong>Simplicity is Key: </strong>Avoid being overly clever. Suno vocalists struggle with significant variations in line length and words with ambiguous pronunciations (e.g., “read,” “live”).</p><h3>Music</h3><h4>Creation</h4><p><strong>Minimal Prompting:</strong> Start simple (e.g., “k-pop, female vocals, uptempo”) to understand Suno’s baseline output.</p><p><strong>Prompt Engineering: </strong>Ask a helper AI (ChatGPT 5.1 or Claude Opus) to draft a Suno prompt.</p><p><strong>Find Good Examples: </strong>Review successful prompts/descriptors on Suno for inspiration.</p><p><strong>Leverage the latest Suno Model: </strong>Use the v5 Suno model for a noticeable quality difference.</p><p><strong>Power User Mode: </strong>As you learn, become more prescriptive with your prompts and experiment with advanced settings, such as Weirdness and prompt adherence levels.</p><p><strong>Accept the Grind: </strong>Getting any song is easy; <em>getting a reasonably good song requires significant iteration.</em></p><h4>Iteration and Editing</h4><p><strong>Remix: </strong>Once you are 75% or more satisfied, use remixing to achieve minor variations. Try varying the prompt slightly for each remix.</p><p><strong>BPM:</strong> Tweak the speed without changing the pitch if the tempo is slightly off.</p><p><strong>Shorten Sections: </strong>Cut down overly long intros or outros.</p><p><strong>Regeneration/Lyrics: </strong>Avoid regenerating sections or replacing lyrics in an existing track, as they rarely match the original. If a section is problematic, consider remixing the entire song with new lyrics.</p><h4>Finalize</h4><p><strong>Good Enough:</strong> Know when to stop. Unless you have professional software to split and edit tracks, there will likely be minor flaws that the AI cannot fix.</p><h4>Remaster</h4><p>You have options depending on your time &amp; level of commitment.</p><p>Auto-Remaster: Use an auto-remaster program (e.g., Suno Remaster or Midea with DistroKid) as a final step before distribution. I tried both and ultimately found I wasn’t happy with either.</p><p>Custom Remaster: Have ChatGPT hand-hold you through remastering in GarageBand or similar. Explain what you don’t like about what you’re hearing in your own words “too quiet, too mushy, etc.” Ask ChatGPT to help you make a mastering template you can apply to each song in your album. This took me hours, but resulted in music I actually wanted to listen to.</p><h3>Album Art</h3><p><strong>Prompt Generation: </strong>Ask ChatGPT to write a DALL-E prompt, feeding it your setlist and song lyrics for context.</p><p><strong>Consistency: </strong>Request DALL-E to maintain consistent colors, themes, fonts, characters, or designs across all song covers, attaching any pre-approved images.</p><p><strong>Reverse Engineer Examples You Like: </strong>Upload an image you like and ask ChatGPT to describe it in artistic detail, then use that description as your DALL-E prompt.</p><p><strong>Iterate: </strong>Refine the prompts with DALL-E and ChatGPT until the art is satisfactory.</p><h3>Ethics and Distribution</h3><p><strong>Follow Guidelines: </strong>Adhere to all guidelines from your music distributor regarding AI-generated music.</p><p><strong>Financial Disclaimer: </strong>Uncheck all boxes that allow you to monetize the music, acknowledging that the models are trained on the work of other artists.</p><p><strong>Declaration: </strong>Declare that the music is AI-generated on all platforms that require it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2183e33c49c1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Practical AI & LLM Use Cases Across the Software Development Lifecycle]]></title>
            <link>https://emekdahl.medium.com/practical-ai-llm-use-cases-across-the-software-development-lifecycle-ca1d59abccee?source=rss-3b9a117391dd------2</link>
            <guid isPermaLink="false">https://medium.com/p/ca1d59abccee</guid>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[large-language-models]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[software-engineering]]></category>
            <dc:creator><![CDATA[Emily Ekdahl]]></dc:creator>
            <pubDate>Tue, 13 May 2025 23:03:04 GMT</pubDate>
            <atom:updated>2025-05-14T09:31:40.071Z</atom:updated>
            <content:encoded><![CDATA[<p>AI isn’t just a hype cycle. It’s quietly becoming a co-pilot across the entire software development lifecycle. From architecture to incident response, large language models (LLMs) are showing up in our workflows to save time, reduce friction, and help us focus on the work that matters.</p><p>Here’s a practical, opinionated roundup of how software engineers use LLMs today. These aren’t theoretical use cases. I’ve done them or seen teams use them to deliver faster within the past few years.</p><h3>Planning &amp; Requirements Gathering</h3><p><strong>Research &amp; Decision Support</strong></p><p>Use AI as a research assistant to get overviews of tools, frameworks, or architectures. It helps cut through noise and spot trade-offs faster when making technical recommendations.</p><p><strong>Brainstorming &amp; Ideation</strong></p><p>LLMs are great for sparking ideas about feature scope, edge cases, and naming. They often surface questions you didn’t think to ask.</p><p><strong>Generate User Stories for Bulk Upload</strong></p><p>Do you need 15+ user stories in a structured format? Just give the LLM a template. It can generate a CSV for bulk upload into Jira, Linear, or whatever tool your team uses.</p><p><strong>Estimations</strong></p><p>Have AI scan your notes, acceptance criteria, or definition of done to flag anything you’ve missed that could derail your estimates.</p><h3>Design &amp; Architecture</h3><p><strong>Drafting One-Pagers, PDDs, and ADRs</strong></p><p>LLMs are great at launching templated documents. You’ll still want to refine and deepen the content but rarely start from scratch.</p><p><strong>Summarizing &amp; Answering Questions from Long or Messy Docs</strong></p><p>Parsing new systems can be painful, between poor library documents and dense academic papers. Pair an LLM (or RAG setup) with your documentation to navigate it faster.</p><p><strong>Generating Code Samples for Spikes or Design Discussions</strong></p><p>Do you need to show how a proposed interface might look? LLMs can sketch it in pseudocode or your preferred language.</p><p><strong>Drafting Process Flows or Diagrams</strong></p><p>Tools like Mermaid, Whimsical, and even some whiteboard apps let you describe a system and auto-generate diagrams — LLMs get you 50 — 80% of the way there.</p><h3>Implementation</h3><p><strong>AI PR Reviewer</strong></p><p>Set up an LLM to summarize pull requests, estimate review time, and flag high-risk changes. This offloads some of the “obvious” comments from senior engineers so they can focus on deeper review.</p><p><strong>AI-Enabled IDEs</strong></p><p>Pair program with AI in your IDE. Review its suggestions like you would an overconfident intern — eager, helpful, but not always right.</p><p><strong>Creative Prototyping</strong></p><p>If your LLM has access to your codebase or design library, it can help rapidly prototype new features via natural language prompts.</p><p><strong>API Usage &amp; Code Patterns</strong></p><p>Ask for examples of interfaces or patterns in context. Bonus points if your IDE can also search live documentation.</p><p><strong>Anticipate Edge Cases &amp; Generate Test Boilerplate</strong></p><p>Tell an LLM the behaviors you want to cover, and it can output structured, boilerplate tests to get you started.</p><p><strong>Debugging Assistance</strong></p><p>LLMs can help make sense of convoluted stack traces, or even pressure-test your hypotheses about what’s going wrong.</p><p><strong>Synthetic Data for AI Evaluations</strong></p><p>Simulating complex user behavior? LLMs can help generate synthetic inputs and edge-case scenarios to evaluate how your system might respond in the wild.</p><h3>Deployment &amp; Ops</h3><p><strong>Config &amp; Infrastructure as Code</strong></p><p>Translate natural language into working CI/CD configs or Kubernetes manifests. Use it to prototype, compare variations, or catch typos early.</p><p><strong>Regression Detection &amp; Results Analysis</strong></p><p>Have the LLM summarize logs or traces and look for themes. The right setup can even correlate logs with metrics or alerts.</p><p><strong>Flag Security &amp; Compliance Risks</strong></p><p>AI-enabled PR reviewers can highlight secrets, insecure settings, or risky changes before they hit main.</p><p><strong>Don’t Repeat Incidents</strong></p><p>Hook your LLM into Incident retros and postmortems. It can flag PRs that resemble patterns that previously caused incidents.</p><p><strong>Ops Bot for Self-Service Access</strong></p><p>If you’ve got good RBAC and policy management, set up a chatbot to automate tool access approvals based on user roles.</p><h3>Documentation &amp; Knowledge Sharing</h3><p><strong>Inline Documentation</strong></p><p>Auto-generate docstrings or comments to lower the activation energy and keep the code well-documented.</p><p><strong>Summarize Incidents → Create On-Call Runbooks</strong></p><p>Use an LLM to pull patterns from incident notes and assemble playbooks for future troubleshooting.</p><p><strong>Automate Release Notes</strong></p><p>Point your LLM at recent tickets and PRs to generate a first draft of release notes — structured, scoped, and ready to polish.</p><p><strong>Defend Focus with an FAQ Bot</strong></p><p>Build a bot with access to chat logs, project docs, and tickets. Let it field common onboarding or how-to questions so engineers stay in flow.</p><h3>Communication &amp; Collaboration</h3><p><strong>Audience-Aware Communication</strong></p><p>Run drafts through one LLM for tone and another for grammar. I almost always have AI rewrite messages for executives to keep things brief and strategy-focused.</p><p><strong>Meeting Summaries with Action Items</strong></p><p>Let an AI note-taker handle documentation so you can stay present in the conversation.</p><p><strong>Onboarding Bots</strong></p><p>Train a bot on your onboarding materials. Let it remind and guide new teammates through setup and orientation.</p><p><strong>Summarize Wall-of-Text Messages</strong></p><p>We all know that colleague. Don’t wade through 14 paragraphs — ask the LLM to extract key points.</p><p><strong>Decode Confusing Messages</strong></p><p>When comms are unclear or vague, get a second opinion before asking follow-ups. Sometimes the AI catches what you missed.</p><p><strong>Simulate Objections</strong></p><p>Test your proposal against predictable pushback. Ask the LLM to raise objections and suggest responses.</p><p><strong>Translation</strong></p><p>Use LLMs to cross language barriers when reading blog posts, documents, or papers, especially in global teams.</p><p><strong>Sentiment Tracking</strong></p><p>Let AI flag sentiment dips in team chats or customer messages. It’s not perfect, but it’s a solid early warning system.</p><h3>Project Management &amp; Team Operations</h3><p><strong>Triage &amp; Prioritization</strong></p><p>LLMs can help tag, sort, and triage tickets, especially when they can pull context from previous similar issues.</p><p><strong>Roadmapping &amp; Scheduling</strong></p><p>Ask AI to break a goal into epics, estimate timelines, or generate a Gantt chart for stakeholders.</p><p><strong>Risk &amp; Dependency Mapping</strong></p><p>Translate ticket “depends on” fields into a clean matrix you can drop into your status deck or cross-team sync.</p><p><strong>Offload Executive Functioning</strong></p><p>If you’re stuck on a task, ask the AI to break it into sub-tasks to help you build momentum.</p><p><strong>Retro Analysis</strong></p><p>Bulk-analyze retrospective notes and extract themes or discussion starters for the next session.</p><p><strong>Stakeholder Updates</strong></p><p>Ask an LLM to generate the first draft of a project update by pulling from tickets, PRs, and changelogs.</p><h3>Wrapping Up</h3><p>The best part of working with LLMs is how quickly they adapt to your workflow. The more context you give them, like templates, patterns, past documents, the more they can take off your plate.</p><p>If you’re using AI in a way I didn’t cover, I’d love to hear about it. Drop it in the comments or message me! Let’s keep leveling up how we work.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ca1d59abccee" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Lint Your Jupyter Notebook With Python Black]]></title>
            <link>https://emekdahl.medium.com/how-to-lint-your-jupyter-notebook-with-python-black-70a7deb45734?source=rss-3b9a117391dd------2</link>
            <guid isPermaLink="false">https://medium.com/p/70a7deb45734</guid>
            <category><![CDATA[python]]></category>
            <category><![CDATA[jupyter-notebook]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[how-to]]></category>
            <category><![CDATA[programming]]></category>
            <dc:creator><![CDATA[Emily Ekdahl]]></dc:creator>
            <pubDate>Sat, 29 Jan 2022 17:39:09 GMT</pubDate>
            <atom:updated>2022-01-29T18:33:54.554Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cftnqRFR6hfr9iFXCo2mEg.png" /></figure><p>First, be sure that you are working in a virtual environment so that you’re managing packages and dependencies safely! If you’re not, check out my article on <a href="https://emekdahl.medium.com/how-to-set-up-your-python-virtual-environment-for-machine-learning-2cfad0b680af">how to set up virtual environments for your Jupyter Notebook.</a></p><p>Next, install <a href="https://github.com/psf/black">Python Black</a> for Jupyter Notebooks.</p><pre><strong>pip3 install black[jupyter]</strong></pre><p>Note that if you’re using <a href="https://ohmyz.sh/">zsh</a> you may need to escape the brackets with a backslash!</p><pre><strong>pip3 install black\[jupyter\]</strong></pre><p>Once you’ve installed it, remember you might need to source your zsh profile again or open a new terminal window.</p><pre><strong>source &lt;your_virtual_environment_name&gt;/bin/activate</strong></pre><p>Then, run this command to lint your notebook:</p><pre><strong>black path/to/your/file.ipynb</strong></pre><p>When linting is complete, you should see this message:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GR-2nTyOgmjuxYxWDGDZEQ.png" /></figure><p>I just love seeing the sparkle cake emoji! Don’t you?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=70a7deb45734" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Use String Interpolation in Jupyter Notebook Markdown Cells]]></title>
            <link>https://emekdahl.medium.com/how-to-use-string-interpolation-in-jupyter-notebook-markdown-cells-8bbe894847ee?source=rss-3b9a117391dd------2</link>
            <guid isPermaLink="false">https://medium.com/p/8bbe894847ee</guid>
            <category><![CDATA[how-to]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[jupyter-notebook]]></category>
            <dc:creator><![CDATA[Emily Ekdahl]]></dc:creator>
            <pubDate>Wed, 07 Apr 2021 13:17:56 GMT</pubDate>
            <atom:updated>2021-04-07T13:17:56.070Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*N2CH-nmAvdBQoZsn3qrBbw.jpeg" /><figcaption>Photo by <a href="https://unsplash.com/@danielfazio?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Daniel Fazio</a> on <a href="https://unsplash.com/s/photos/string?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></figcaption></figure><p>As a data professional, you may want to discuss results in a Jupyter Notebook markdown cell without hard coding the actual values in case they change. This is where string interpolation comes in!</p><p>String interpolation is a common practice in software development and is used to insert a variable and make it appear as text like the information around it.</p><p>Unfortunately, string interpolation does not work as intuitively in Jupyter Notebook markdown as it does in software development. Great news, the path to string interpolation in markdown cells has been forged for you!</p><p>Here are the steps I took to string interpolate text in a markdown cell.</p><p>First, install a Jupyter Notebook extension.</p><pre><strong>pip install jupyter_contrib_nbextensions</strong></pre><pre><strong>jupyter contrib nbextension install --user</strong></pre><p>Then, set it to enable python-markdown.</p><pre><strong>jupyter nbextension enable python-markdown/main</strong></pre><p>Start your Jupyter Notebook and create a variable or variables. Then, use the double curly braces to string interpolate your variable and execute the cell.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZZrEfWMgHXN3_Y2IE1sRsw.png" /></figure><p>Hope this helps!</p><p>One small caveat, when I attempted to convert my notebook to PDF, the values did not display.</p><p>Special thanks to those who <a href="https://stackoverflow.com/questions/52812231/print-variable-in-jupyter-notebook-markdown-cell-python">shared this information</a> on stack overflow.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8bbe894847ee" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Set Up Your Python Virtual Environment for Machine Learning]]></title>
            <link>https://emekdahl.medium.com/how-to-set-up-your-python-virtual-environment-for-machine-learning-2cfad0b680af?source=rss-3b9a117391dd------2</link>
            <guid isPermaLink="false">https://medium.com/p/2cfad0b680af</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[how-to]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[python]]></category>
            <dc:creator><![CDATA[Emily Ekdahl]]></dc:creator>
            <pubDate>Sat, 03 Apr 2021 13:04:27 GMT</pubDate>
            <atom:updated>2021-04-03T14:14:41.783Z</atom:updated>
            <content:encoded><![CDATA[<h3>How to Set Up Your Jupyter Notebook and Python Virtual Environment for Machine Learning</h3><p>If you’re like me, you have to remind yourself of the commands each time you set up a new project. I’m writing it down here for me and for you!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YYhCnkeZCgWHiKUbs0Z-vA.jpeg" /><figcaption>image credit: <a href="https://unsplash.com/@sadswim">https://unsplash.com/@sadswim</a></figcaption></figure><h3>The first time</h3><p>Open your bash or zsh profile.</p><p>For example, I use zsh.</p><pre><strong>zshconfig</strong></pre><p>If you used bash and vim, it would look something like this.</p><pre><strong>vim ~/.bash_profile</strong></pre><p>Hit i to go into insert mode.</p><p>Export the path to your project as a variable to make your life easier.</p><pre><strong>export MY_ML_PROJECT_PATH=&quot;$HOME/fancy_new_ml_project&quot;</strong></pre><p>Hit ESC to exit insert mode.</p><p>Save and close your bash or zsh profile with :wq</p><p>Make a directory with that variable you created above.</p><pre><strong>mkdir -p $MY_ML_PROJECT_PATH</strong></pre><p>Check your versions of python and pip.</p><pre><strong>python3 --version</strong></pre><pre><strong>python3 -m pip --version</strong></pre><p>Upgrade pip if you need to do so.</p><pre><strong>python3 -m pip install --user -U pip</strong></pre><p>Install python virtual environment package with pip.</p><pre><strong>python3 -m pip install --user -U virtualenv</strong></pre><p>Navigate to your project directory and create your virtual environment.</p><pre><strong>cd $MY_ML_PROJECT_PATH</strong></pre><pre><strong>python3 -m virtualenv my_env_name_of_choice</strong></pre><p>Activate your virtual environment.</p><pre><strong>source my_env_name_of_choice/bin/activate</strong> # on Linux or macOS<br><strong>.\my_env_name_of_choice\Scripts\activate</strong>  # on Windows</pre><p>Install these packages and/or your ML packages of choice.</p><pre><strong>python3 -m pip install -U jupyter matplotlib numpy pandas scipy scikit-learn seaborn</strong></pre><p>Name and register the virtual environment to Jupyter Notebook.</p><pre><strong>python3 -m ipykernel install --user --name=ml</strong></pre><p>When you start your new project inside Jupyter Notebook, make sure the new virtual environment you named and registered is selected!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/358/1*Lm3iWfKw99Kw68sSUR3-fQ.png" /></figure><h3>Every time</h3><p>Whenever you need to navigate to your project after setup, follow these steps.</p><pre><strong>cd $MY_ML_PROJECT_PATH<br>source my_env_name_of_choice/bin/activate</strong> # on Linux or macOS<br><strong>.\my_env_name_of_choice\Scripts\activate</strong>  # on Windows</pre><p>Have fun!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2cfad0b680af" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Resolve SSL: CERTIFICATE_VERIFY_FAILED Error on Mac]]></title>
            <link>https://emekdahl.medium.com/how-to-resolve-ssl-certificate-verify-failed-error-on-mac-77de2a46ef2b?source=rss-3b9a117391dd------2</link>
            <guid isPermaLink="false">https://medium.com/p/77de2a46ef2b</guid>
            <category><![CDATA[python]]></category>
            <category><![CDATA[debugging]]></category>
            <category><![CDATA[apple]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Emily Ekdahl]]></dc:creator>
            <pubDate>Sun, 15 Nov 2020 13:47:02 GMT</pubDate>
            <atom:updated>2020-11-15T13:47:02.724Z</atom:updated>
            <content:encoded><![CDATA[<p>Really excited to get a new demo or project up and running? Great! 🎉</p><p>Cloned/started the repo? Awesome! 🤓</p><p>Got this error? Oh no! 😞</p><pre>URLError: &lt;urlopen error [SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate</pre><p>Try this command, but make sure you put whatever version of Python you’re using in place of <strong><em>3.8:</em></strong></p><pre>open /Applications/Python\ <strong>3.8</strong>/Install\ Certificates.command</pre><p>Several of the solutions I saw recommended this command, however, when I ran the first one, it kicked off this second one for me:</p><pre>pip install --upgrade certifi</pre><p>Happy hacking!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=77de2a46ef2b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Insert Images in Jupyter Notebook]]></title>
            <link>https://emekdahl.medium.com/how-to-insert-images-in-jupyter-notebook-5abcb3064f0a?source=rss-3b9a117391dd------2</link>
            <guid isPermaLink="false">https://medium.com/p/5abcb3064f0a</guid>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[python]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[jupyter-notebook]]></category>
            <dc:creator><![CDATA[Emily Ekdahl]]></dc:creator>
            <pubDate>Wed, 11 Nov 2020 17:19:33 GMT</pubDate>
            <atom:updated>2020-11-11T17:19:33.592Z</atom:updated>
            <content:encoded><![CDATA[<p>Sometimes it is just easier to insert a picture than to explain with text or code comments. This brief article shows you how to insert images into Jupyter Notebooks!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/760/1*jGrudpOSJBq2W5hwmB_RwA.jpeg" /><figcaption>Gonzalez (2016)</figcaption></figure><h3>Markdown</h3><p>First, set your cell to markdown!</p><h4>Local Images</h4><p>Assuming your file structure looks like this:</p><pre>/<br>+-- your_jupyter_notebook_file.ipynb<br>+-- img<br>    +-- image_file_name_here.png</pre><p>Insert your image like this:</p><pre>![Title](img/image_file_name_here.png)</pre><h4>Web Image</h4><p>For web images, you can simply insert the url:</p><pre>![Random Unsplash Image](<a href="https://source.unsplash.com/random">https://source.unsplash.com/rando</a>m)</pre><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5abcb3064f0a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Five Shortcuts for Improved Productivity in Jupyter Notebooks]]></title>
            <link>https://emekdahl.medium.com/five-shortcuts-for-improved-productivity-in-jupyter-notebooks-590826259a7f?source=rss-3b9a117391dd------2</link>
            <guid isPermaLink="false">https://medium.com/p/590826259a7f</guid>
            <category><![CDATA[jupyter-notebook]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Emily Ekdahl]]></dc:creator>
            <pubDate>Tue, 22 Sep 2020 11:58:12 GMT</pubDate>
            <atom:updated>2020-09-22T11:58:12.618Z</atom:updated>
            <content:encoded><![CDATA[<p>Data Scientists can and should emulate software engineers’ obsession with productivity hacks. Keyboard shortcuts are a quick and dirty way you can become better at your job. (Disclaimer: these shortcuts are specifically for Mac Users!)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3Oj0dhFEpxVdp0-Gl_q4MA.png" /></figure><p>If you know one thing, know how to <strong>execute your cells</strong> without clicking “Run.” This command also automatically creates the next cell:</p><blockquote>SHIFT + ENTER</blockquote><p>Since proper documentation is key, the second thing to master is <strong>switching between code and markdown cells</strong> without reaching for your mouse or trackpad.</p><p>Markdown:</p><blockquote>ESC + M then hit ENTER</blockquote><p>Code:</p><blockquote>ESC + Y then hit ENTER</blockquote><p>Sometimes you end up with code you no longer want or need. <strong>Deleting old or dead code</strong> is one of the great joys of software engineering.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*f8fouKXBDIwCiNg_o4iWsg.png" /></figure><p>(Shout out to all you vim users, this may look familiar to you!)</p><blockquote>ESC + D + D</blockquote><p>There’s a <strong>wonderful world of shortcuts to explore</strong> in Jupyter Notebooks and you can have the rest of them at your fingertips by using this keyboard shortcut:</p><blockquote>ESC + H</blockquote><p>Thanks for reading! Hope this helps you become more productive and efficient while working in Jupyter Notebooks!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=590826259a7f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Aspire]]></title>
            <link>https://emekdahl.medium.com/aspire-eec0b3e9cb85?source=rss-3b9a117391dd------2</link>
            <guid isPermaLink="false">https://medium.com/p/eec0b3e9cb85</guid>
            <category><![CDATA[vuejs]]></category>
            <category><![CDATA[javascript]]></category>
            <category><![CDATA[ruby]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[web-development]]></category>
            <dc:creator><![CDATA[Emily Ekdahl]]></dc:creator>
            <pubDate>Thu, 15 Jun 2017 01:10:13 GMT</pubDate>
            <atom:updated>2017-06-15T01:10:13.005Z</atom:updated>
            <content:encoded><![CDATA[<p>Aspire is a single page personal finance app for digital natives. Inspired by the crisis of student loan debt, Aspire was developed to help millennials leverage their most important asset — time. Written with Vue JS framework, the app dynamically updates values, graphs, and dates whenever the user enters new info without having to re-render the page. On the back end, Ruby runs on a Rails framework and produces a JSON API to save and retrieve users’ financial data from the postgreSQL database.</p><p>Here’s the tour: check out the YouTube video below!</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2F_7y-usiBWok%3Ffeature%3Doembed&amp;url=http%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3D_7y-usiBWok&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2F_7y-usiBWok%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="640" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/2cacf414b68494cb64c4054ef52c7318/href">https://medium.com/media/2cacf414b68494cb64c4054ef52c7318/href</a></iframe><h3>Highlights</h3><ol><li>Using <a href="http://momentjs.com/">Moment.js</a> to create a cascading timeline — this was a feature that I felt was missing in many other financial calculators. There are calculators for debt repayment, saving, and investing but I hadn’t yet seen one that combined the three to project the full experience start to finish.</li><li><a href="http://materializecss.com/">Materialize</a> front-end framework made the parallax design clean and easy. Many of the financial calculators I’ve run across look old or outdated. I believe having a more modern appearance adds to the credibility of some apps.</li><li>Responsive charts that danced whenever the user adjusted the data, courtesy of <a href="http://vue-charts.hchspersonal.tk/">Vue-Charts</a> wrapper on the Chart.js library. This is the fun interactive part of the financial calculator, the one that encourages the user to try out different scenarios.</li><li>Automagic calculations without clicking a button or re-rendering a page, thanks to <a href="https://vuejs.org/">Vue.js</a>! Easily the bulk of the time I spent on the app was in converting personal finance calculations to JavaScript and then getting them to all work together. Many of the functions were dependent on or called other functions. Getting them all working together was super rewarding!</li></ol><h3>Challenges and Opportunities</h3><ol><li>Multiple types of debt — In the end, I didn’t have time to write the function that would take in any number of different debts with their own interest rates and minimum payments, and calculate the payoff date. Though this was my initial plan, making the entire path (debt repayment =&gt; emergency fund =&gt; investments) was more important to me. If I had more time, I would build out this functionality. It would require more tables in the database and a more sophisticated debt repayment algorithm.</li><li>Adjust income or savings over time- Realistically, you’ll probably get a raise at some point (hopefully!) The current calculations only have one entry point for income and savings toward financial goals. I did that on purpose to streamline the app; however, it would be more interesting to game out other complex scenarios like a 3% raise every few years, bonuses, etc. Also, it would be nice to be able to say, for example, I will save 30% of my income and aggressively paydown debt and save an emergence fund, then I will (increase/decrease) my savings toward financial goals by XX% in the financial independence step.</li></ol><p>Thanks for checking out my web app! I had a lot of fun building it and hope that you enjoyed it, too!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=eec0b3e9cb85" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Adapting Materialize Form Component for Ruby on Rails]]></title>
            <link>https://emekdahl.medium.com/adapting-materialize-form-component-for-ruby-on-rails-70f9ba068bf9?source=rss-3b9a117391dd------2</link>
            <guid isPermaLink="false">https://medium.com/p/70f9ba068bf9</guid>
            <category><![CDATA[materialize]]></category>
            <category><![CDATA[ruby-on-rails]]></category>
            <category><![CDATA[html5]]></category>
            <category><![CDATA[web-development]]></category>
            <category><![CDATA[css]]></category>
            <dc:creator><![CDATA[Emily Ekdahl]]></dc:creator>
            <pubDate>Sun, 19 Mar 2017 21:46:49 GMT</pubDate>
            <atom:updated>2017-03-19T21:54:43.949Z</atom:updated>
            <content:encoded><![CDATA[<p>When using Materialize forms with Ruby on Rails, the default form doesn’t play well with .html.erb.</p><p>Here’s a basic form component on the <a href="http://materializecss.com/forms.html">Materialize website</a>, with <strong>&lt;form class =”col s12&quot;&gt; </strong>and <strong>&lt;input id=”variable_name_here”&gt; </strong>:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/479/1*gJEwsocmZ9DARs1CejhJvg.png" /></figure><p><strong>Where things go terribly wrong:</strong></p><ol><li>Trying to mash up the Ruby’s <strong>&lt;% form_tag ‘/controller_name’, method: :post do %&gt;</strong> with Materialize’s <strong>&lt;form class=”some_class_here”&gt;</strong></li><li>Thinking that <strong>id=”variable_name”</strong> in the default Materialize form will work the same way as Ruby’s <strong>name=”variable_name”</strong> in the <strong>&lt;input&gt;</strong> field.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/481/1*G5n13SNkNECzHabqM7AH8g.png" /></figure><p>If you use Materialize default form class, or try to combine it with Ruby’s form_tag, you get form-ception and your data never reaches its destination. Furthermore, when you submit your form it doesn’t seem to go anywhere because the form class keeps it from executing the the form_tag’s post request and sending you on to your next page or success condition.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/475/1*Ee9d3jrJ89Z01C3zCenSNA.jpeg" /><figcaption>don’t be like Leo, use Ruby’s form_tag and ditch Materialize’s form class</figcaption></figure><p>This is what it looks like with <strong>form class</strong> swapped for Ruby’s <strong>form_tag</strong> and <strong>name=””</strong> swapped in for <strong>id=””</strong>:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/560/1*2ZPR1OMNXeRQVvUNzo6BUw.png" /></figure><p>If you use this strategy, your form will post to your database and you can proceed to your next page or success condition. Good luck and happy coding!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=70f9ba068bf9" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>