Skip to content

Hello! I'm Ayo

I'm Ayo Alfonso — a software engineer. Now, at Souk. I build production AI systems — tool use, model routing, evals, and the guardrails that make autonomy safe to ship.

SUMMARY

  • CTO at Souk, building AI agents for partnerships
  • Before that, fraud risk at Trudenty — ML risk scoring and privacy-preserving verification
  • Earlier, engineering at Github, Hotels.ng, Union bank of Nigeria
  • Since 2017 I've built software for startups, banks and marketplaces in Lagos and London
  • Based in London, United Kingdom
  • I built the first grassroots scout marketplace product in Africa
  • I write about tech and engineering on Medium ↗
Ayo Alfonso

PROJECTS

Hermione Agent
Hermione Agent

Lovable for mobile games. An agent that turns a plain-English brief into a shipped Flutter game. Smart model routing: cheap models triage, frontier models take over when the work needs real reasoning. It picks its own verification checks as it builds — and is barred by design from touching live player state

AI AgentsGamingTooling
Tymbah
Tymbah ↗

AI meeting assistant that automatically generates summaries and action items from your meetings

ProductivitySaaS
CapitalRollup
CapitalRollup ↗

A DeFi app written in Flutter, simplifies investing for users to DCA into stable coins

DeFiBlockchainFinance
Verbalyst
Verbalyst

An immersive AI language learning app written in React native

EducationEdTechLive Learning

AI RESEARCH

Fast Inference under Real-Time Constraints
Fast Inference under Real-Time Constraints

Research into optimizing accent conversion models for real-time conversation, focusing on reducing latency while maintaining quality.

AIResearchML Ops
SHAP + LIME: A Practical Hybrid for AI Model Explainability
SHAP + LIME: A Practical Hybrid for AI Model Explainability

My Msc. project under Azita pourghasem deep dives into how SHAP and LIME explainability frameworks work, asking how we can combine the best of both worlds

AIExplainability
Multi-Layer Fine-Tuning for Specific Accent Groups
Multi-Layer Fine-Tuning for Specific Accent Groups

Research into developing fine-tuning strategies that allow targeted improvements for specific accent groups while maintaining general performance.

AIResearchML Ops

WHAT I'M READING

RECENT POSTS