Available · Staff+ · Co-founder · ML Consulting

Staff Full-Stack ML Engineer · AI/Fintech Systems

When AI behavior, platform reliability,and product clarity must hold simultaneously —you need someone who has built all three.

The system has to work at 2am.

LIVE IN PRODUCTION

End-to-end ML systems

SabiScore sustains 99.9%+ uptime (Prometheus · 90-day window) — ensemble XGBoost, LightGBM, and CatBoost inference with 45% MTTD improvement over reactive alerting baseline.

DECISIONS DOCUMENTED

Architecture visible

Every tradeoff — Chosen, Over, Because — is in the case study. Not just outcomes. Reasoning at every level, legible without clicking.

ZERO-DOWNTIME DESIGN

Graceful by default

Health checks, idempotent job queues, circuit breakers, and environment-scoped limits built in from deployment one — not bolted on after an incident.

FULL OWNERSHIP

Feature to inference

Feature engineering through FastAPI inference to the Next.js frontend. One engineer, full stack — no handoff latency, no translation loss, no waiting.

Selected Work

Work that shipped

End-to-end AI and fintech systems. Each ships with documented architecture decisions and a monitored production deployment.

ML Platform · ObservabilityLIVE

SabiScore

Production ML prediction and self-monitoring platform for live decision windows.

Ensemble credit and prediction scoring (XGBoost, LightGBM, CatBoost) with real-time output quality monitoring. Alerts engineers the moment a model begins degrading — before any user is affected. 99.9%+ uptime (Prometheus · 90-day window). ~30% inference latency reduction via query optimisation and Redis caching. 45% MTTD improvement over reactive alerting baseline.

Architecture Decision

CHOSEN

FastAPI + Redis Pub/Sub for inference serving

OVER

Synchronous REST with database polling

BECAUSE

Sub-50ms event fan-out at sustained load with dead-letter recovery — impossible with polling under concurrent sessions

Constraint: Ensemble inference must complete in <120ms p99 at peak load with no model warmup on cold start.

FastAPI
XGBoost
LightGBM
CatBoost
Redis Pub/Sub
Prometheus
Grafana
PostgreSQL

Live demo GitHub Case study

Privacy Infrastructure · BlockchainWIP

Hashablanca

Multi-chain encrypted data transfer with ZK-proof regulatory compliance.

Backend for secure, verifiable transfer of sensitive data across Ethereum, Polygon, BSC, and StarkNet simultaneously. Multi-GB AES-256 payloads with automatic retry and full transaction audit records. ZK-SNARK proof pipelines for GDPR-compliant regulatory reporting. 90% integration test coverage. Zero data-loss record throughout.

Architecture Decision

CHOSEN

ZK-SNARKs for regulatory proof generation

OVER

Centralised compliance oracle with full data exposure

BECAUSE

GDPR mandates minimum data disclosure — ZK proofs validate compliance without revealing payload contents to the verifier

Solidity
ZK-SNARKs
IPFS
Ethereum
Polygon
BSC
StarkNet
Node.js

GitHub Case study

Compliance Platform · FintechCASE STUDY

TaxBridge

Nigerian SME tax filing from 4 hours to 15 minutes — NRS-integrated, audit-ready.

Full tax compliance workflow automation for Nigerian small businesses — VAT, withholding tax, and annual returns. PostgreSQL RLS isolates each tenant at the database level. Real-time calculations under <150ms at load. Idempotent BullMQ job queue ensures no submission is ever double-processed — even through mid-request server failure. Hash-chained immutable audit trail. 95% test coverage.

Architecture Decision

CHOSEN

PostgreSQL Row-Level Security for multi-tenancy

OVER

Application-layer tenant filtering

BECAUSE

NRS audit scrutiny demands proof that tenant data cannot cross-contaminate — RLS enforces this at the database engine level, not the application layer

Fastify 5
PostgreSQL 15 RLS
Redis 7
BullMQ
GraphQL
Prisma
TypeScript

Case study

Background

The engineer

Backend engineer and platform architect with four years of independent product and consulting work — and over a decade building critical data infrastructure within Nigeria's federal public sector.

Non-CS degree, FUTO Environmental Technology 2011. Credibility built through production engineering, four cloud certifications (AWS, GCP, OpenJS Node.js, PostgreSQL), and 15+ merged upstream contributions to XGBoost and scikit-learn — third-party, verifiable signals.

At UBEC — the Universal Basic Education Commission — working in the Planning, Research and Statistics department. Building the ETL pipelines and dashboards that consolidate national education data from all 36 state offices and feed federal budget allocation decisions.

Currently consulting across fintech and SaaS teams in West Africa and Europe — auditing backend systems, resolving reliability issues, and building internal tooling alongside CTOs and senior engineers.

Lagos, Nigeria — open to remote.

Certifications

AWS Certified Developer

Dec 2023

GCP Associate Cloud Engineer

Aug 2023

OpenJS Node.js Services (JSNSD)

May 2024

PostgreSQL 14 Associate

Mar 2024

Open source

Node.js · PostgreSQL

pg-tenant

Enforces strict database-level data separation between clients. One tenant's records are completely invisible to another even when application-layer bugs exist. Used in production by fintech teams.

Fintech · Compliance

audit-chain

Generates a cryptographically linked audit log where every entry is mathematically bound to the previous one. Makes retroactive edits or tampering instantly detectable. Built for NRS and GDPR trails.

AI · DevOps

node-debug-llm

Streams live system logs and traces to an AI model, which returns a ranked plain-English list of likely root causes — compressing hours of manual incident investigation into minutes.

THE CUT

Engineering in depth

Long-form breakdowns of shipping decisions, platform constraints, and what continues to hold when production stops being polite.

Featured1 published

Building SabiScore: Architecture Decisions in a Resource-Constrained Production System

How ensemble model selection, Redis caching, and graceful fallback shaped a production ML API under uneven infrastructure conditions.

12 Nov 20252 min readML ENGINEERINGFASTAPIPRODUCTION SYSTEMS

Read article

View all writing (1)

Open — responding within 48hrs

Start a conversation

Available for Staff+ roles, technical co-founding, and scoped ML consulting where reliability is a requirement rather than a nice-to-have.

Staff+ / Principal

Distributed systems · ML platforms · API infrastructure

Available for Staff+ and Principal Backend roles at AI-native fintech and product companies. Four years of independent platform work with zero-downtime design as a baseline constraint.

Technical Co-Founder

Pre-seed to Series A · Africa/emerging markets

Four years shipping production platforms from zero. Infrastructure, ML, and compliance architecture through funding rounds. The system should outlast the seed deck.

ML Consulting

Model deployment · Observability · Performance

Inference serving, monitoring pipelines, and latency reduction. SabiScore approaches 30% inference latency reduction (Redis caching + query optimisation). Engagements scoped to specific problems with measurable outcomes.

scardubu@gmail.com LinkedIn GitHub +234 803 388 5065

The engineer

Backend engineer and platform architect with four years of independent product and consulting work — and over a decade building critical data infrastructure within Nigeria's federal public sector.

Lagos, Nigeria — open to remote.