About Fikra API
We are closing the access gap. AI inference optimized and hosted locally in Kenya, built specifically for developers locked out of global payment infrastructure.
Cheaper Inference
$2/1M tokens vs. $15/1M globally.
African Languages
Localized support for Swahili, Sheng, and regional contexts.
Local Rails
Top up with local options directly. No credit card required.
Support Line
Talk directly to the person writing the code.
// Why Fikra Exists
"I kept watching developers across East Africa hit the same wall — Stripe not supported, $20 minimum too steep, no card at all. The tools to build exist. The access doesn't. That's the gap Fikra API is closing."
James Miano
Founder & CTO, Roniki Systems · Nairobi, Kenya
Low-Level Weight Optimization
To achieve our 10x pricing advantage without burning cash, we optimize the execution model directly. James is one of the few ML engineers globally developing custom pipeline optimization for Ternary Weight (1.58-bit) LLM architectures.
By quantizing model layers into simple values of -1, 0, and 1, we eliminate memory-intensive matrix multiplications entirely. Hardware runs fast, power consumption drops, and you save.
// Quantize matrix weights to ternary states (-1, 0, 1) fn quantize_1_58bit(weight: f32, scale: f32) -> i8 { let normalized = weight / scale; if normalized > 0.5 { 1 } else if normalized < -0.5 { -1 } else { 0 } } // Inference matrix addition replacing multiplication for weight in model_layers { match weight { 1 => output += input, -1 => output -= input, 0 => continue, } }
// The Founder
James Miano
- ▪ Self-taught AI/ML engineer focused strictly on inference optimization and scaling.
- ▪ Designing integrated Groq-accelerated backend architectures.
- ▪ Advocating for localized infrastructure ownership in the global South.
// Contact Direct
Get in Touch
Whether you're looking for pricing optimizations, architectural insights, or want to discuss enterprise integrations, you can reach out directly.
We guarantee absolute transperancy. Follow along and support our building in public journey.