Ian Too

80 posts

Ian Too banner
Ian Too

Ian Too

@ianktoo

San Jose, CA Katılım Aralık 2012
415 Takip Edilen401 Takipçiler
Sabitlenmiş Tweet
Ian Too
Ian Too@ianktoo·
Kaggle and HuggingFace have thousands of datasets. But what if none of them are yours? I built DataForge for exactly that moment. When you know the domain, the tone, the source, and you just need the data shaped right for fine-tuning. Point it at a URL. It crawls, chunks, generates synthetic Q&A / instruction / conversation pairs, scores quality, and exports straight to HuggingFace, Kaggle, or local files. "Just use an existing dataset." Sure, until your use case is specific enough that nothing fits. Experimental. Open-source. Needs a quick LLM setup (tool walks you through it). Would love for people to try it and break it. 👉 github.com/ianktoo/data-f… pip install llm-web-crawler
English
0
0
0
270
Ian Too
Ian Too@ianktoo·
📍Coyote Point Recreational Area, San Mateo, CA No LLMs out here ☺️, feels good. #AWalkInThePark
Ian Too tweet mediaIan Too tweet mediaIan Too tweet mediaIan Too tweet media
English
0
0
1
132
Ian Too
Ian Too@ianktoo·
Hey @Microsoft @Windows, why can't we search through our clipboard history (Win+V)? The search bar only looks for emojis. Having a real search button for copied text would save so much time! Please consider adding this. Pretty please 🥺 #Windows #Productivity
English
0
0
2
45
Ian Too
Ian Too@ianktoo·
This is how tafsiri works with @DarajaAI Project: github.com/ianktoo/tafsiri
Ian Too tweet media
Ian Too@ianktoo

Came across @DarajaAI and their African language models while building Crea (an emergency response app). That stopped me. Language is a barrier in emergencies, a wrong translation can be worse than none. So I built tafsiri. An eval pipeline that runs on @DarajaAI API as translation engine. Right now, it translates into Swahili, Yoruba, Amharic (rides on Daraja). Evaluates each output confidence (Daraja's own confidence), back-translation (asks Daraja the reverse translation) and LLM-as-judge (any frontier large language model). Not just a translation. A score you can trust. It's < 5 days old. Open source. No stats yet. Try it out! github.com/ianktoo/tafsiri

English
0
2
2
129
Ian Too
Ian Too@ianktoo·
Came across @DarajaAI and their African language models while building Crea (an emergency response app). That stopped me. Language is a barrier in emergencies, a wrong translation can be worse than none. So I built tafsiri. An eval pipeline that runs on @DarajaAI API as translation engine. Right now, it translates into Swahili, Yoruba, Amharic (rides on Daraja). Evaluates each output confidence (Daraja's own confidence), back-translation (asks Daraja the reverse translation) and LLM-as-judge (any frontier large language model). Not just a translation. A score you can trust. It's < 5 days old. Open source. No stats yet. Try it out! github.com/ianktoo/tafsiri
English
1
4
7
1K
Ian Too
Ian Too@ianktoo·
@DarajaAI @KelvRoman Thanks! So helpful! Here's some feedback: - Would be helpful to have it in the dashboard (API Management) too right where the API Keys live, or have a link in the footer. I have more helpful feedback (I think), please lemme know how I should share 😊
English
1
0
2
27
Kelvin Roman
Kelvin Roman@KelvRoman·
The first decision I made as a founder wasn't about code or models or funding. It was the name. Daraja. Bridge. In Swahili. A bridge connects what was separated. A bridge doesn't choose sides. A bridge exists for everyone who needs to cross. Monday, you'll see what that means. daraja.ai #AIforAfrica
English
3
6
27
1.3K
Ian Too
Ian Too@ianktoo·
Publishing on app store is extreme sports 😫 🏗 @ridewithhalo
English
0
0
0
27
Ian Too
Ian Too@ianktoo·
The AMD Developer Hackathon is live on lablab.ai right now. I might join. Mostly because I want to understand how AI runs on AMD hardware, and there's no better way to learn than to build something under pressure with people who already know. Sharing in case it's relevant to anyone here.
English
0
0
0
53
Ian Too
Ian Too@ianktoo·
Working on Crea.ai. Goal is to run a finetuned model on the edge. Not there yet. What I do have runs locally and works better than expected. Just don't ask it to get creative under pressure. Bad idea. Finetuning is hard at the start and hard at the end. The middle is just curiosity. I love the middle.
Ian Too tweet media
English
0
0
0
31
Ian Too
Ian Too@ianktoo·
F450 + Pixhawk 1 + Raspberry Pi 4 + Marvelmind indoor GPS + Tuya T5-E1... See where I'm going with this? 🏗
Ian Too tweet mediaIan Too tweet mediaIan Too tweet media
English
0
0
2
1.3K
Ian Too
Ian Too@ianktoo·
This paper digs into the gaps in Kenya’s National Address System and compares it with global approaches. Solving this would unlock huge value across logistics, emergency response, and GIS. Worth reading: frontiersin.org/journals/susta…
English
0
0
0
26