John (@John_TheAnalyst) header
John (@John_TheAnalyst) profile picture

ML Engineer · Building AI systems for Agriculture, fintech, DeFi & public safety · From Nigeria, working globally.

Remote
Joined August 2022
Photo by John (@John_TheAnalyst)Photo by John (@John_TheAnalyst)Photo by John (@John_TheAnalyst)Photo by John (@John_TheAnalyst)Photo by John (@John_TheAnalyst)Photo by John (@John_TheAnalyst)Photo by John (@John_TheAnalyst)Photo by John (@John_TheAnalyst)Photo by John (@John_TheAnalyst)Photo by John (@John_TheAnalyst)
AI companion apps can shut down and take your companion with them. Ownling is an AI pet that owns itself. It thinks on NEAR AI's sealed hardware, pays for every thought from its own NFT, and wakes up with our servers off. @NEARProtocol #NEARHackathon
2
2
164
The ML angle The project that made me an ML engineer. It started with a call with my boss. The idea: a system that predicts commodity prices in Nigeria. I said yes. Then I hung up and thought: how do I even do this? 🧵
So my journey to ML actually started last year. Last year I was getting into blockchain analysis on Dune. I was reading everything: Ethereum, Solana, Uniswap, token metrics. Honestly? It was overwhelming. I was fluent in Web2 data. Web3 felt like a different language. Thread🧵
1
1
38
Today, @AgroLinking the system covers 18 commodities across 12 states in all 6 geopolitical zones, with forecasts from the next day to 6 months ahead. It has a live website, an API, price alerts, and an accuracy page that shows how right and how wrong we are.
1
3
What it taught me: ML engineering isn’t about the fanciest model. It’s mostly data, pipelines and honest testing. Remember Aerodrome, where I learnt a number means nothing without a baseline? That’s what saved this project. Knowing when not to use ML is part of the job.
4
92 NEAR DAOs paid out $99.8M last year. For 68% of it, nothing says what it was for. Tickmark reads a Sputnik DAO or Trezu treasury from NEAR, sorts every payment and lists what nobody explained. Built at NEAR HACKATHON. 04-john.github.io/Tickmark/ @NEARProtocol #NEARHackathon
1
63
This is true, I've done a lot of learning this past week than I think over for this year because of her
Every day i wake up and watch a new interview by her Every single interview of her has so many things to take and so many things to learn
1
1
65
92 NEAR DAOs paid out $99.8M last year. For 68% of it, nothing says what it was for. Tickmark reads a Sputnik DAO or Trezu treasury from NEAR, sorts every payment and lists what nobody explained. Built at NEAR HACKATHON. 04-john.github.io/Tickmark/ @NEARProtocol #NEARHackathon
2
61
How: every DAO payment is approved by an on-chain vote, so Tickmark dates it by that vote, prices it that day and checks the balances add up. The council then approves the books with one vote. Code: github.com/04-John/Tickmark Treasurers: what would your DAO need to use this?
1
30
AI companion apps can shut down and take your companion with them. Ownling is an AI pet that owns itself. It thinks on NEAR AI's sealed hardware, pays for every thought from its own NFT, and wakes up with our servers off. @NEARProtocol #NEARHackathon
2
1
3
98
Try it in your browser, no wallet needed: Site: ownling.onrender.com Code: github.com/04-John/Ownling You can also download it on Android and Windows: ownling.onrender.com/downloa… What would you want a pet that can't be taken from you to remember about you?
31
Today on how i got my fist job in On-Chain Analysis as someone who was upskilling towards ML The job was ClankFun, and how did I lande my first paid job in blockchain analytics. It didn’t come from an application. It came from a Telegram message 🧵
Tomorrow I'll be talking about how I got my firtclient in the world on blockchain while working on getting into ML
1
1
105
I didn’t do it alone. I worked with Gideon, and three weeks later we delivered and got paid. The bigger win came after: a year on, we’re still building together. Sometimes the client is the job, but the person you meet along the way is the real return.
1
14
That closes my blockchain chapter: Uniswap, Aerodrome, OdinBot, ClankFun. Four projects that taught me how data is built, cleaned, defined and delivered. And they all pointed in one direction: machine learning. That’s where the story goes next.
15