@dataRecode

Data Analytics. Data Science. Data Visualization. Geospatial Maps. Statistics. for more contact us at [email protected]
www(dot)dataRecode(dot)com

[@dataRecode @datarecode]

Africa's 2021 electric power generation: just 3% of global output. This choropleth map and charts highlight the urgent need to boost capacity and diversify energy sources to drive AI innovation and GDP growth. #PowerGeneration #AfricaEnergy #rstats

[@dataRecode @datarecode]

Communicating results visually is part science part art. Ultimately, collaboration between humans & AI is best for effective visual communication/design. Here is an essay resulted from working with AI on real world project. #dataViz #rstats #datascience

[@dataRecode @datarecode]

Built and tested CreditSense, a full-stack turnkey prototype that combines deterministic rules engines with secure local LLMs to auto-generate professional credit risk memos in seconds.Successfully running multiple reports in parallel. #CreditRisk #FinTech #AI #rstats

[@dataRecode @datarecode]

📢💫Messy data = hallucinating AI. The real problem isn’t the model, it’s the dirty data we feed it.

Just dropped a new post on why data cleaning should be treated as real engineering + how we’re fixing it with #typecleanR. 👇https://t.co/zQrOy9i32x
#DataScience #AI #Rstats

[Abiyu Giday @abiyugiday]

💫Just added a starter vignette to #typecleanR 👇
📏Roadmap highlights:
* Factor coercion → detect categorical text & convert to factors
* Integer coercion → distinguish whole numbers vs decimals
* Currency & percentage coercion → strip $, €, ETB or convert “45%” cleanly
* Boolean & empty string coercion → broader logical detection + consistent NA handling
* Scientific notation coercion → safely parse 1.23e4

Messy inputs → clean types → stronger analytics & safer #LLM orchestration. #rstats

[Abiyu Giday @abiyugiday]

📢 📦New R package alert: typecleanR
I’ve just released an early version of typecleanR on GitHub! A tidyverse‑friendly toolkit for cleaning coercion types in a data.

It provides helpers for:
* Numeric coercion → safely convert text to numbers.
* Date coercion → strict conversion when all values match.
*Logical coercion → handle TRUE/FALSE/T/F consistently.
* Lenient datetime parsing → flexibly parse mixed formats, drop bad values to NA.
👉 It’s live here: https://t.co/18MmeFSWeX
Install with:

devtools::install_github("abiyug/typecleanR")

💡 I’d love for the community to test run it, break it, and share feedback before I polish docs and add CI. Your input will help shape the package into something robust and useful for messy Excel/CSV workflows. #rstats #R

[@dataRecode @datarecode]

Abiyu Giday
@abiyugiday
·
Most AI tools feel generic because they don’t know your data.I We turned 3,300 scattered datasets into a structured catalog with proper tags and rich context. The difference is huge. Read linkedIN 👇: https://t.co/MYOOWrvsyb

#rstats #analytics #LLM #RAG

[@dataRecode @datarecode]

Happy (belated) International Women's Day! 🌟

A random #ethiopia|n women image processed via #metrR for data-driven distortion, visualized in #ggplot2 as layered contours #InternationalWomensDay #DataArt #IWD2026 #rstats (ps: code is in the alt code)

[@dataRecode @datarecode]

Why does parallel zip metadata extraction in R go 100× slower? Overhead kills small jobs! Sequential wins on 27 archives (~136 MB). Benchmark > guess.

We turn data chaos into speed at dataRecode. hit us up: [email protected]
#Analytics #HighPerformanceComputing #BigData

[@dataRecode @datarecode]

👇💫 #rtidbits

[Abiyu Giday @abiyugiday]

💫👇 Data wrangling pro tip
Skip arrange() + slice(1) when you want the first or last row per group.

Use slice_min() and slice_max() instead. These two function can directly give you the alphabetically smallest or largest value per group.
#rstats @RosanaFerrero #tidydatatuesday

[@dataRecode @datarecode]

direct link to the package. https://t.co/s8yygGtMYz

[@dataRecode @datarecode]

Quickly peek inside ZIP files without extracting them! Check filename & other metadata across archives, and more 💍. Perfect for data dumps, backups, model checkpoints etc.Give it a try:
remotes::install_github("abiyug/zipAuditR")
Repo: https://t.co/r3L0X5o6EM #rstats

[@dataRecode @datarecode]

📢When you try to embed a YouTube video using an iframe in #RMarkdown and knit a website, you may get Error 153. This is due to a recent YouTube security policy change regarding local file origins. Here is a Gist that lets you work around that: 👇https://t.co/C4Eh0gM2Z3
#rstats

[@dataRecode @datarecode]

Running a daily experiment: same tight prompt to 8 diverse LLMs + predictive models. Their reasoned verdicts get lined up side-by-side. The magic? The disagreements that’s where the real insight hides. Read more about it here👇https://t.co/rYBtBycJoQ
#AI #LLMs #rstats

[@dataRecode @datarecode]

Cloud LLMs are powerful. But local models aren’t going anywhere. Legal, private, and domain specific needs demand flexible AI infrastructure. That’s why we built a customisable Web UI for LLMs w/ streaming support, role-based agents, and hybrid model switching. #LLM #rstats

[@dataRecode @datarecode]

Screenshot of an early prototype of our LLM web UI. It supports streaming and works with both local and cloud-hosted models. A web UI for LLMs offers a browser-based interface for prompt-response interaction — no terminal required. Powered by #shinychat, #ellmer, and #flexdash.

[@dataRecode @datarecode]

Checkout our latest blog on Chaining R workflow with locally hosted LLMs over local area network. Read more detail on the blog plus code here https://t.co/YzNZ5hgQKQ #tidyllm #rstats #ai

[@dataRecode @datarecode]

Generate polished, audience-specific #HTML automated reports from the same data, for board, CXO, or shareholders, without touching #CSS. ##bslib + #LLMs let you style, preview, and automate effortlessly so you focus on insights. #rstats #markdown

[@dataRecode @datarecode]

Smarter AI, Safer Setup: Enriching LLM Responses with Local Data in R. We built a local workflow using R + Ollama to enrich LLM responses with private documents🔗 Full walkthrough + code [https://t.co/TutixwaLcI] #rstats #AI #RAG #ReproducibleResearch

[@dataRecode @datarecode]

And .... we're back! [https://t.co/uiCA3QKf5X] is now live and accessible! Thanks for your patience!

[Abiyu Giday @abiyugiday]

📢 Website Access Update for Our Ethiopian Visitors

We’ve heard from many of you that our website [dataRecode (dot) com] isn’t currently loading in #Ethiopia. We’re aware of the issue, and it appears to be related to broader network restrictions affecting GitHub-hosted sites.

We’re actively working on a solution to ensure smooth access for everyone, including exploring alternative hosting options that are fully accessible within Ethiopia.

In the meantime, if you need anything — product info, support, or updates — feel free to reach out directly here or via DM. We’re here and ready to help.

Thank you for your patience and continued support. We’ll keep you posted as we resolve this!

[Abiyu Giday @abiyugiday]

📢 Website Access Update for Our Ethiopian Visitors

We’ve heard from many of you that our website [dataRecode (dot) com] isn’t currently loading in #Ethiopia. We’re aware of the issue, and it appears to be related to broader network restrictions affecting GitHub-hosted sites.

We’re actively working on a solution to ensure smooth access for everyone, including exploring alternative hosting options that are fully accessible within Ethiopia.

In the meantime, if you need anything — product info, support, or updates — feel free to reach out directly here or via DM. We’re here and ready to help.

Thank you for your patience and continued support. We’ll keep you posted as we resolve this!