Data Stack - The Roundup #1
5 projects from 5 builders working across data, ML, and AI. Real work, straight from the source.
Intro
If you follow my notes, you know I have been spending time connecting with data folks here on Substack. Reading their work, exploring their projects, tools, research, and everything in between. At some point during a conversation with Ame_data scientist , something clicked. These people are building genuinely interesting things. Why not bring them together in my publication and get their work in front of more eyes? That is how this started.
Data Stack is an initiative to bring developers, engineers, and practitioners working in Data, ML, and AI together in one place. The goal is simple: build a real community around the work, not just the discourse.
The Roundup is one part of that. Builders submit their own projects, in their own words. A tool, an app, a dataset, a course, a research paper. Something they built and something useful to people in this space. What gets submitted is what gets published, no editing, no paraphrasing.
The first edition features 5 projects from 5 contributors including me :) Different backgrounds, different problems, different tools. Take what is useful, follow who interests you, and if you are building something worth sharing, the next edition is open for submissions.
🛠️ The Projects
01 - SnapTask
By YAZ - Mobile Architect & Independent Developer
🔗 SnapTask
I built SnapTask to solve “app-switching fatigue” and input latency for builders who need to capture thoughts the second they happen. When you’re deep in data pipelines, model training, or coding, opening a heavy cloud-synced notes app completely breaks your cognitive flow, and the idea is often gone before the UI even loads.
SnapTask is a lightweight Android utility designed to eliminate this friction entirely while respecting developer prerequisites. It injects a minimal input shortcut directly into your notification bar. You swipe down, type, and save with zero load time.
Crucially, it features built-in natural language date parsing. Instead of forcing you to fiddle with clunky UI calendars or time pickers, it automatically extracts deadlines from how you naturally think and type, whether that’s “assignment in 2 days” or “build script next monday”.
Most importantly, it is 100% offline-first and private. Your data stays entirely on your device with zero external cloud dependencies, tracking, or telemetry. It’s built purely to minimize input friction and protect your cognitive overhead, letting you offload ideas instantly and get right back to building.
02 - DFM
By Martha - Independent Developer
🔗 No link was shared by Martha.
DFM is an AI project based on the Phi 3.5 Mini language model. I completely rewrote its code architecture, adding a unique multi-agent loop that I invented myself to increase its creative intelligence and novelty.
Basic small language models usually have very low depth of thinking, low creativity, and tend to give surface-level answers. DFM, on the other hand, has a much greater depth of thought, which allows it to engage very deeply with user prompts. It can invent novel ideas and create genuinely sophisticated responses that usually only larger AI models can produce.
More than that, it can invent its own topics and ask unexpected questions — something that even models like Claude or ChatGPT 5 usually don’t do.
It’s important to understand that DFM is not a system designed for useful or highly logical answers. It has messy grammar, and it can melt down or hallucinate a lot. There are already many language models for logical and practical responses — Claude and ChatGPT exist specifically for that.
The original goal was to create an AI that can think deeply, draw unexpected parallels, and constantly surprise the user with its answers.
I’ve never really seen an AI respond the way DFM does, which is interesting. It’s essentially a novel invention that doesn’t currently have an analogue.
03 - 2026 Roadmap to Becoming a Data Analyst
By Brien Grant | Data Analyst - Data Analyst
The reason for me to make the road map is because people were reaching out to me on Linkedin for how I switched from a teacher to a data analyst. So, I thought there is a demand for this. Secondly, my growth on Linkedin just keeps on going up. Currently at 9,400 followers over there. It took me 5 weeks to make it. I had a plan of what I wanted to hit. First Excel/Google sheets, then I moved to Dashboard/Visualization. Those are like the core skills people want. From there I abstracted it to SQL, and Python. After having all those technical skills down. Now people need to understand what they are look at which lead to Domain Knowledge. The order is Spreadsheet, Visualization, SQL, Python and Domain Knowledge.
My only challenge was time with being a new father and more responsible at work. So just time factor. Making the content wasn’t hard because it was like being a teacher again. Setting up a lesson plan and then having a homework. The other reason is I don’t think I have left the classroom and I still want to help and teach people. Just now in the domain of data.
04 - Blog2Video App
by Arslan Shahid - AI Engineer
I almost spent $30,000 turning my blog into videos. My blog was working. Clients found me through long-form content. But traffic plateaued. Every algorithm was pushing video. So I tried to make the switch.
First attempt: hired video editors. $300–$1,000 per video. 50+ posts. You do the math. Not happening. Second attempt: AI video generators. Generic stock footage. Robot voice. Didn’t sound like me, didn’t use my actual content. AI slop at $20–50 a video.
So I built Blog2Video.
It scrapes your actual blog. AI programs the video in Remotion — animated diagrams, code blocks, bullet reveals — rendered as React components. No AI image models. No hallucinated visuals. Just your content, your voice, your brand.
The key insight: don’t pay AI to generate pixels. Use AI to write code that builds the video programmatically.
50 posts converted. $0 to video editors. $30K saved. Now it is making me money.
05 - KI Reads
By Nazan | NextInData 😊
🔗 KI-Reads
I follow data and AI blogs scattered across different platforms. Some are on Substack, some on Medium, some on personal sites. Keeping up with all of them was taking more time than actually reading them. It was time consuming to check platforms if something new had been posted.
I wanted a simple solution. No new app to check, no platform deciding what I see first, no algorithm pushing popular content over the stuff I actually care about.
So I built KI-Reads.
How does it work? You send us your list of blog and newsletter URLs, we fetch the latest posts and deliver them in one clean email every Sunday morning to your inbox with the latest posts from the past 7 days, organized by category. Any platform, any RSS feed, your list, your rules.
No app to download, no account to create, no tracking, no ads.
That wraps up the first edition of The Roundup.
These are real projects built by people actively working in the data and AI space. If something caught your attention, follow the builder and dig deeper. The best way to support this kind of work is to engage with it directly.
The next edition is already in the works. If you have something worth sharing with this community, submissions are open.
See you in the next one!