Obinna Iheanachor

Do you want to become a world-class data analyst in just 6 months? I’ve been there, and I know how overwhelming it can be to learn data analysis from scratch. That’s why I’ve created a 6-month roadmap of free high-quality resources that will teach you everything you need to know, from statistics and Excel to Python and cloud computing. In this thread, I’ll share with you my ultimate roadmap and how you can download it for free. Let’s get started! 👇

We're thrilled to welcome @DataSenseiObi to TDS! Click below to explore his comprehensive walkthrough on designing a streamlined, rapid document-extraction system. towardsdatascience.com/from-4-weeks-t…

I replaced £8,000 of manual work with ~$15 in API calls. 4,700 PDFs. 45 minutes. 96% accuracy. Published on @TDataScience 👇 towardsdatascience.com/from-4-weeks-t… Here’s the architecture (and why GPT-5 wasn’t the answer) 🧵

@FNakabugo Glad you found the article insightful Faridah!

@DataSenseiObi This article is very well explained. As someone new to the field, I was able to clearly understand the problem the approach solves. I especially appreciated the discussion around why different approaches were chosen and how they align with stakeholder needs.

If you’ve built extraction pipelines: What edge case broke yours at scale?

What actually broke the system: → revision tables mistaken for current values → grid letters (A, B, C) read as revisions → PDFs rotated with incorrect metadata None of this showed up in small tests.

I spent 20 mins pitching an AI project My manager asked: "How much time does this save us?" I didn’t have an answer. Proposal died Second time: one slide. 4 lines. Approved! Same project. Same tech. Different framing Here’s the exact framework ↓ youtu.be/XmeSiUUE-cc?si…

Full breakdown + a one-slide pitch template you can screenshot and send to your manager. 👉 Watch Episode 3: youtu.be/XmeSiUUE-cc?si… This closes Arc 1 of my Production AI series. Arc 2: we build.

And when they ask “what if it’s wrong?”: Don’t say “it’s 95% accurate.” Say “every result is verifiable.” The first invites doubt. The second invites trust.

First time I pitched an AI project, I talked about embeddings for 20 minutes. My manager: "How much time does this save us?" I didn’t have an answer. Proposal died. Here’s what I learned 🧵

SQL failed. Not because the data was missing. Because it didn’t understand the question. RAG did instantly. This changes how we query data. 2-min demo 👇 youtu.be/hu5XweKRsh8?si…

I asked the SAME question on the SAME data. SQL: 0 rows RAG: exact answer + citations Nothing was missing. SQL just needed the *exact wording*. RAG understood the intent. This is the shift most data people are underestimating: 7 data skills that map directly to AI 🧵