Three questions to ask any AI vendor before you sign:
1. Do we own the weights?
2. Where does our data live?
3. Does the bill grow with usage?
Babel-Forge: you own the weights, your data stays home, no usage meter. Ask everyone else the same.
@DarajaAI, AI for Africa
Daraja retweeted
Before you sign with any AI vendor, ask three questions:
1. Do we own the weights?
2. Where does our data live?
3. Does the bill grow with usage?
Most vendors fail at least one. Ask early, while you still have leverage.
The Africa-and-AI story is always a "gap" to bridge: access, catch-up.
Bridge it and you are still renting tools built elsewhere.
Africa does not need to catch up on AI. It needs to own it. Its own models, its own data, its own soil.
@DarajaAI, AI for Africa
Daraja retweeted
Every headline about Africa and AI uses the same word: gap.
Retire it.
A gap makes Africa a tenant, renting tools built elsewhere, on someone else's terms. Close it and you still own nothing.
Africa needs to catch up to Frontier Labs but most importantly, it needs to own it.
Cloud was supposed to end vendor lock-in. AI rebuilt it and charges by the word.
A rented model means your core intelligence on infra you do not control, running the same brain as every competitor.
Babel-Forge: a model that is yours, on your hardware, no meter.
Daraja retweeted
The cloud was supposed to end vendor lock-in.
AI rebuilt it, and now charges you by the word.
Your core intelligence, on infra you do not control, priced by a meter you do not set, same brain as every competitor.
Babel-Forge: a model that is yours, on your hardware, no meter.
Own your core.
"Any model" is not an answer. Babel-Forge trains and hands over, on your own infrastructure:
reasoning models in your languages
domain models on your documents
voice models for non-typists
code models on your stack
assistants grounded in your private data
You own all of it.
When someone asks what Babel-Forge actually trains for them, they deserve specifics. So here is the real menu, and every one of these is a model we train and hand over to run on your own infrastructure:
A reasoning model that thinks in your languages, the official ones and the ones your customers actually code-switch into.
A domain expert, trained on your documents, your policies and your products, so it reasons inside your field instead of guessing at it.
A voice model for the people who will never type a query, built for low literacy and real accents.
A code model trained on your stack and your own codebase, not a generic autocomplete.
A retrieval-grounded assistant wired into your private knowledge, answering from your truth and not the open internet.
YOUR model shaped around your business. Not a rental you share with everyone else.
Daraja retweeted
When someone asks what Babel-Forge actually trains for them, they deserve specifics. So here is the real menu, and every one of these is a model we train and hand over to run on your own infrastructure:
A reasoning model that thinks in your languages, the official ones and the ones your customers actually code-switch into.
A domain expert, trained on your documents, your policies and your products, so it reasons inside your field instead of guessing at it.
A voice model for the people who will never type a query, built for low literacy and real accents.
A code model trained on your stack and your own codebase, not a generic autocomplete.
A retrieval-grounded assistant wired into your private knowledge, answering from your truth and not the open internet.
YOUR model shaped around your business. Not a rental you share with everyone else.
Put your mind at ease by having us create your own intelligence - yes model and weights
daraja.ai/babel-forge
Daraja retweeted
This is a great example of why organizations should own their own intelligence.
At @DarajaAI, we build custom made LLM tailored to your organization’s specific wants, needs and preferred capabilities.
To request a quote or walkthrough, check out our Babel-Forge program daraja.ai/babel-forge
We were able to fine tune an open weight model (Qwen3.5-9B) on proprietary internal data at @coinbase, and it out performed frontier models, at much lower cost and speed, for a specialized task (fraud prevention in this case).
I think we'll see more of this pattern inside companies. The frontier models are amazing generalists. But increasingly you'll see specialist models out perform on more narrow tasks, either working on teams, or by themselves.
We're making it easy to fine tune models internally for our employees. We even have a skill for it in our internal harness that makes it easy to spin these up and start training them:
"/fine-tune-and-serve-llm I want to fine-tune Qwen for payments risk review. help me get started"
I imagine we'll have some backend services that could be replaced entirely by this new architecture.
Babel-Forge, from Daraja.
We train a model on your data. You deploy it inside your own walls. You own it outright.
No API subscription. No black box. Nothing leaves your building.
Daraja retweeted
Every bank and ministry I talk to wants AI and cannot use it. One question ends the meeting: where does our data go.
So we built Babel-Forge. We train a model on your data. You run it inside your own walls. You own it outright.
Nothing leaves your building.
Daraja retweeted
True, responsible stewardship will determine our future. This is why at @DarajaAI we prioritize building rigorous safety checks, clear testing protocols and transparent oversight into every stage of development. Our goal is to ensure AI serves humanity’s need while minimizing unintended consequences.
techcabal.com/2026/09/24/cou…
@TechCabal
Daraja retweeted
Here are some common questions our stakeholders and users have asked us before and after adding Babel to their workspaces.
daraja.ai/slack
Daraja retweeted
I asked Babel-Asaro what are the benefits of having an African AI model that can and how it, being a Reasoning Model, can benefit the African continent. This was its response:
Today we’re unveiling the next generation of African Intelligence: Babel-Asaro.
Babel-Asaro is the first multilingual reasoning AI built specifically for African languages - the first model to combine chain-of-thought reasoning with support for multiple African languages at once.
daraja.ai/chat
@TechCrunch, @TechCabal
More than half a billion people speak these twelve African languages.
AI used to translate them. Babel-Asaro reasons in them, step by step, in the language asked.
More than half a billion people speak these twelve languages.
For all of them, AI has only ever translated. Carry your words into English, think in English, send an answer back.
Babel-Asaro reasons in their languages instead. Step by step, in the tongue they think in.
The thinking machine finally speaks theirs too.
Daraja retweeted
More than half a billion people speak these twelve languages.
For all of them, AI has only ever translated. Carry your words into English, think in English, send an answer back.
Babel-Asaro reasons in their languages instead. Step by step, in the tongue they think in.
The thinking machine finally speaks theirs too.
