Job Description
You'll make didii actually smart. Voice intent parsing, fraud detection, Naija-context understanding — the AI is half the product.
What you'll do
- Train and ship ML models for fraud, spend insights, and conversational intent
- Tune LLM prompts and guardrails for money actions
- Build evaluation harnesses so we don't ship hallucinations
- Partner with engineering to deploy models at low latency
What we're looking for
- 3+ years applied ML/AI — not just notebooks, real users
- Comfortable with PyTorch, HuggingFace, modern LLM tooling
- You've shipped a model to production and watched it behave under load
- You understand why a 1% fraud false-positive rate is catastrophic at scale