Skip to main content
This page exists so anyone evaluating Guild for the OPay Scholars 2026 National Innovation Challenge can verify the alignment in two minutes.

​
The problem we set out to solve

A mid-sized African nation is experiencing high youth unemployment and a fragmented informal economy. Design an intelligent economic system powered by data and AI that connects informal traders, job seekers, and financial services in one ecosystem. What does it look like, how does it work, and how does it scale?
Forty-one million Nigerians work in the informal economy. They earn every day. They spend every day. They keep families running. But to a bank’s underwriting model, they don’t exist — because there is no transaction trail a system can read. The fragmentation runs across three sides at once. Job Seekers can’t prove their work history. Employers lose ten to fifteen percent of their labour cost to ghost workers and middlemen. Financial institutions, even the ones who want to lend, have no way to underwrite people whose income is real but invisible. Connecting any two of those sides is incrementally useful. Connecting all three is the bet.

​
The five capabilities at Guild’s core

Guild was built to deliver five core capabilities for informal-economy inclusion. Each one is a real surface in the product today.

1. Identify and digitally onboard informal traders and job seekers

Three onboarding rails so nobody is excluded by their device:
  • Voice (offline phone call). A bricklayer in Modakeke on a Tecno Spark dials Guild. Tola, our voice agent, asks three questions in Nigerian English: name, location, what they do. Account created.
  • WhatsApp. A trader at Oje Market sends “I want to join.” A five-step Flow inside the chat completes the signup.
  • Web. Phone + OTP. No password.
Same identity record across all three. Tier 1 is phone + OTP. Tier 2 adds profile + payout bank. Tier 3 adds BVN, NIN, and address verification — at which point the user is eligible for the full partner consent flow.More detail: Sign up, Trader onboarding, Employer onboarding.

2. Match job seekers to opportunities using AI

The four factors are exactly the four the brief names: skill, location, language, economic context.Our matching model is a learning-to-rank tree ensemble (lambdarank objective). It runs the moment any job posts, scores every plausible candidate, and surfaces the top names to the employer’s dashboard while pinging the top candidates on the rails they use (WhatsApp, SMS, outbound voice call).
  • Skill: trade category match, skill-tag overlap, experience level fit, past similar-job completion count.
  • Location: distance in km, state match, area-within-city match, distance-decay weighting.
  • Language: preferred language vs spoken languages, with strict-match and partial-match scoring.
  • Economic context: standing tier, on-time rate, average review, recent activity, tenure, repeat-employer count.
Inference takes under 100ms for a typical candidate pool. Retrained nightly on the day’s new completed-job and review data.More detail: Matching model, Find work.

3. Connect traders and workers to financial services

Credit, savings, insurance, and payments — all on the same ledger, all underwritten on alternative data and behavioural signals rather than traditional credit history.
  • Payments. Live today through OPay. Direct debit from employer to Job Seeker’s wallet to Job Seeker’s bank in seconds.
  • Savings. Rules-based automatic transfers from wallet to a separate savings balance. Goal tracking. Withdraw to bank.
  • Credit. The Partner API surfaces signed snapshots of a user’s standing tier, behavioural signals, and audit ledger to partner financial institutions. They underwrite on what we send. Inventory advances for traders. Tool-equipment loans for skilled Job Seekers.
  • Insurance. Same ledger, future product surface.
The standing score is the alternative-data underwriting signal. A Job Seeker who has completed 38 jobs in 90 days with an average review of 4.7 is creditable. A trader with ₦95k daily inflow velocity across 47 distinct customers is creditable. The data says so.More detail: Standing score, Sales velocity, Partner Portal, Partner API → Snapshots.

4. Learn and improve over time as data flows

Four models, each one calibrated by the platform’s continuous activity.
  • Standing model. Retrained nightly on new ledger rows. The longer Guild runs, the sharper the tier assignment.
  • Matching model. Retrained nightly on the day’s accept/decline/complete signals. Employer approvals and Job Seeker completions are direct training signal.
  • Negotiation engine. Online Thompson-sampling bandit. Posterior updates immediately on every accept or decline. Each (segment × bucket) pair gets sharper with every interaction.
  • Transcription model. Calibrated to Nigerian English, Pidgin, Yoruba, Igbo, Hausa.
The data flywheel is not aspirational. It is the operating mode.More detail: ML layer.

5. Integrate OPay API as a core transactional and data layer

Not optional, per the brief. Not optional in our architecture either.Twenty-three OPay surfaces wrapped in one client. Five fire on every payment story you’ll read on this docs site:
  • Direct Debit (Mandates). Employers link a bank once via the ₦50 confirmation transfer. After that, every job posting auto-funds escrow.
  • Payouts. NIBSS-backed transfers to any Nigerian bank in seconds.
  • Webhooks. Three signature schemes verified, reconcile-guarded against OPay’s authoritative verify endpoint before any wallet credit.
  • Virtual Accounts. Static B2B accounts for traders. Dynamic VAs for invoice-specific collections.
  • Transactional SMS. OTPs, balance alerts, payment confirmations.
The other eighteen cover card linking, direct-payment fallbacks, bills (airtime, data, electricity), and reconciliation. Wired. Production-grade. Every OPay call carries a ULID idempotency key with Postgres-durable + Redis fast-path replay protection. Every OPay call is wrapped in an opossum circuit breaker. Every webhook is verified with crypto.timingSafeEqual before processing.More detail: OPay integration, OPay reference.

​
How we hit each of the four pillars

The OPay Scholars 2026 National Innovation Challenge weighs solutions across four innovation pillars. Guild engages all four — meaningfully, not symbolically. Here’s how each surfaces in the system today.

AI Automation

AI-powered workflows that automate the future of financial technology.
  • Tola, our voice agent. Inbound and outbound calls in Nigerian English with code-switching into Pidgin, Yoruba, Igbo, Hausa. Six tools fire on a single call — sign up, list jobs, get details, apply, update profile, set skills.
  • WhatsApp intent engine. Twenty-three conversational intents across five languages. Voice notes interpreted by our transcription model and routed to the same intent engine as typed messages.
  • WhatsApp Calling. Tap-to-call Tola from inside the chat. SIP bridge with Opus-to-PCMU transcoding. No separate dial.
  • Auto-funded escrow. A job posts, escrow funds, releases fire per shift as attendance is marked. Zero employer taps after submit.
  • Outbound nudges. The matching model surfaces a high-fit candidate; voice-svc places the outbound call.
No human in the loop for any of these workflows.

Use of Data

Predictive, data-driven solutions that protect communities and keep transactions secure in real time.
  • Append-only audit ledger. Every job, every shift, every payment, every webhook is a row. BEFORE UPDATE OR DELETE triggers reject mutation. The history is structurally immutable.
  • Standing model. Four-tier classifier reading the ledger. Recomputed nightly. Surfaces to the dashboard and to the Partner API as a signed snapshot.
  • Matching model. Lambdarank, four factors (skill, location, language, economic context), retrained nightly on accept/decline/complete signals.
  • Negotiation engine. Online Thompson-sampling bandit, per-segment Beta posteriors, updates on every interaction.
  • Webhook reconcile guard. Every credit-bearing webhook is re-verified against OPay’s authoritative endpoint before any wallet credit. Defends against signature-replay even if a secret leaks.
  • Encrypted PII. BVN, NIN, address — envelope-encrypted, AAD-bound to user ULID. A database leak alone gives only ciphertext.

OPay APIs

Smart financial services scaling on OPay’s powerful API stack.
  • 23 surfaces wrapped in one client. Direct debit (mandates), payouts, virtual accounts (static + dynamic), reconciliation, transactional SMS, hosted checkout, card linking, direct-payment fallbacks, bills (airtime/data/electricity).
  • 5 surfaces fire on every payment story. Direct debit, payouts, webhooks, virtual accounts, transactional SMS.
  • Three signature schemes verified on inbound webhooks. Timing-safe comparison every time.
  • Idempotency end to end. ULID keys on every OPay call. Postgres-durable + Redis fast-path. No double debits, ever.
  • Circuit breakers with cached-response fallback on read-only endpoints. Demo-resilience without sacrificing correctness.
Full reference: OPay: complete surface reference.

Financial Innovation

Bold solutions that transform how money moves.
  • Receipts for the people the banks forgot. Every kobo through OPay lands in a ledger a lender can verify. Three months of activity becomes credit identity.
  • Three-way marketplace. Job Seekers ↔ Employers ↔ Banks & financial services. Connecting any two is incrementally useful; connecting all three closes the loop.
  • Escrow-on-post. Employer’s bank is debited at the moment a job posts, not the moment it ends. Funds release per shift as attendance is marked. Zero ghost workers.
  • The Partner API. Consent-gated, signed snapshots. Underwriting on alternative data, not credit-bureau lookups. Inventory advances for traders. Tool-equipment loans for skilled Job Seekers.
  • One ledger, four products. Payments today. Credit through the standing score. Savings through wallet-source rules. Insurance on the same ledger when it ships.

​
How it scales

The brief asks for a path from a 10,000-user pilot to national deployment. Our architecture was designed for it from day one. More detail: Resilience & scale.

​
Three additional commitments the brief implicitly assumes

The brief mentions ethics and viability. We took those seriously. More detail: Security & trust.

Introduction

The Musa story. The thesis. The five rails.

How it works

Adekunle posts a job. Musa applies. Money moves. The score updates. The lender reads.

Architecture

Eight services, sixteen shared packages, the OPay client at the centre.

The Partner Portal

The third side of the connection — banks and financial services reading signed snapshots.