Things I've Shipped
Real problems.
Real solutions.
Still running.
Every project here solved a real problem. Here's the situation, what I built, and what changed because of it.
The Problem
Digital wallet implementations are often fragile under load and poorly documented — causing integration delays and unreliable user experiences. There was no clean, testable reference implementation to build from.
What I Built
A scalable wallet service using Node.js, TypeScript, and PostgreSQL with Prisma ORM. Containerised with Docker, Redis for caching and transaction validation, Jest for unit testing, and Swagger for complete API documentation.
The Outcome
Developer integration time cut by 60% through thorough Swagger documentation. High system reliability confirmed through Jest test coverage. Production-ready and containerised for any deployment environment.
The Problem
HR manually pulled attendance data from biometric devices every month — a slow, error-prone process that created reconciliation headaches and kept non-technical staff dependent on IT for basic data exports.
What I Built
Automated attendance retrieval from ZKTeco biometric devices via SDK integration. Structured ExcelJS report generation in formats HR could immediately use. A self-service dashboard — no IT ticket required.
The Outcome
Manual errors and reconciliation time cut by 80%. HR gained full autonomy over their attendance data. IT dependency for routine exports dropped to zero.
The Problem
Raw attendance data stored in proprietary .attlog formats was inaccessible to HR teams without manual intervention — creating bottlenecks, data entry errors, and slow reporting cycles that impacted payroll and workforce management.
What I Built
An attendance processing system capable of decoding and extracting employee data from .attlog files for automated record handling. Built file upload and parsing workflows to process raw attendance logs and dynamically update structured Excel reports. Backend services for attendance validation, data transformation, and report generation using Node.js and TypeScript.
The Outcome
Operational efficiency significantly improved and manual data entry eliminated. Optimised data processing logic ensured accurate attendance tracking, streamlined reporting workflows, and reliable handling of large attendance datasets.
The Problem
Large, unstructured audio files were unsuitable for direct AI/ML analysis — requiring a reliable way to segment audio into meaningful, consistently-sized chunks while maintaining the security of private audio assets.
What I Built
A system to segment large audio files into meaningful chunks for AI/ML analysis, integrated with Google Cloud Storage for secure handling of private audio files.
The Outcome
Delivered fast, structured audio segmentation while enhancing security and compliance — providing the AI/ML pipeline with clean, ready-to-process audio data at scale.
The Problem
AI models trained on Western-centric datasets fail to capture African language nuance and cultural context. A robust, secure platform was needed to collect multilingual conversation data from native speakers at scale for AI model training.
What I Built
Backend infrastructure for LangEasy — secure user data handling, GCS integration for audio and text assets, API endpoints supporting multilingual data collection workflows, and compliance mechanisms for sensitive language data.
The Outcome
Live product actively powering AI model training. Supports real-time multilingual conversations, personalised learning flows, and cultural data annotation feeding production AI models.
The Problem
Data QA was fragmented — team members working across disconnected tools with no unified audit trail, making it impossible to track data quality at scale for AI model training.
What I Built
A multi-user QA system with role-based access across three stages: Annotate, Audit, and Export. Backend handles concurrent sessions, structured data validation, and exportable quality reports feeding the ML pipeline.
The Outcome
Unified QA process across the entire data annotation team. ML engineers receive clean, audited datasets. Measurably improved the quality and consistency of data fed into AI model training.
The Problem
Running AI inference workloads on resource-constrained edge devices is inherently difficult — requiring optimised service layers, reliable data transmission to cloud services, and fault-tolerant systems that perform consistently in demanding real-world environments.
What I Built
Deployed and optimised AI inference workloads on Raspberry Pi and NVIDIA Jetson Nano/Orin platforms for real-time processing and intelligent automation. Integrated edge hardware with backend APIs and cloud services for secure, efficient data transmission. Designed lightweight, scalable service layers tailored for resource-constrained devices, with performance optimisation and fault-tolerance strategies to reduce latency and ensure consistent operation.
The Outcome
Improved system responsiveness, operational stability, and memory efficiency across edge deployments. Enabled seamless deployment, monitoring, and maintenance of AI-powered embedded solutions through close collaboration across software and hardware workflows.
The Problem
Leave application and approval workflows were unstructured and manual — creating inconsistencies in authorisation, poor visibility into staff availability, and no centralised system for HR to manage workforce coordination effectively.
What I Built
A comprehensive HR leave management platform using Node.js, TypeScript, and a scalable backend architecture. Implemented a multi-level approval and rejection system enabling Team Leads and HR personnel to manage requests through structured authorisation pipelines. Built attendance tracking, employee profile management, and automated birthday notification features. Designed secure RESTful APIs and optimised data handling processes within a cross-functional development team.
The Outcome
Delivered a reliable, user-friendly enterprise HR solution that streamlined leave workflows, improved workforce coordination, and gave HR full visibility over employee leave and attendance data.
Experience
3 years of building
things that work.
From banking data at Sterling to AI infrastructure at Awarri — a track record of delivery across engineering, data, and automation.
Nov 2023 – Present
Full-time
Awarri
Software Developer (Backend)
🏆 Staff of the Year · 2025
Awarri was building AI-driven products that required robust, scalable backend infrastructure — with no existing architecture in place. The team needed an engineer who could own the full backend pipeline from the ground up.
Architect and deploy scalable backend systems powering AI-driven products. Build and optimise data pipelines. Integrate secure authentication mechanisms. Collaborate cross-functionally to debug, refine architecture, and enforce best practices.
Architected and deployed scalable backend infrastructure powering AI-driven products. Built and optimised data pipelines, reducing processing latency and improving retrieval speeds. Integrated Google CORS authentication and secure validation mechanisms, strengthening compliance. Collaborated with cross-functional teams to debug, refine architecture, and implement best practices.
Delivered backend systems aligned with business needs, ensuring high availability and performance. Live AI products shipped to production. Awarded Product Team Staff of the Year 2025 — recognised for consistent excellence, leadership, and cross-functional delivery.
Nov 2023 · Lagos, Nigeria
Full-time
Awarri
Machine Learning Data Annotation Specialist
Awarri's machine learning models required high-quality, large-scale annotated datasets to perform accurately — but lacked a structured annotation process to support reliable model training at scale.
Train machine learning models by conducting accurate, large-scale data annotations. Validate datasets to improve model performance and ensure consistent, reliable project outcomes.
Conducted accurate, large-scale data annotations to train machine learning models. Applied precise validation techniques to maintain data integrity and consistency across annotation batches.
Improved model performance and project outcomes through precise validation, contributing to higher-quality AI training data that directly enhanced the reliability of Awarri's machine learning models.
Aug 2019 – Jan 2020
Internship
Sterling Bank Plc
Data Analyst Intern
Sterling Bank needed analysts who could translate operational data into actionable insights for senior stakeholders — not just report numbers, but interpret them.
Analyse operational data, surface trends, build visual dashboards, and track KPIs that would inform strategic decisions at management level.
Analysed operational datasets to identify workflow efficiency patterns. Built visual dashboards and reports presented directly to senior stakeholders. Developed structured KPI tracking frameworks to improve performance measurement accuracy.
Delivered reports that fed into strategic planning decisions at senior level. Built the analytical foundation — structured thinking, data interpretation, stakeholder communication — that underpins every engineering decision I make today.
Education
🎓
B.Sc. Mathematics
Nnamdi Azikiwe University · Awka, Anambra
2016 – 2021
Certifications & Training
🖥️
HP IT for Business Certificate
HP
💬
Effective Listening & Communication Foundations
LinkedIn Learning
⚡
Full-Stack Developer Trainee
SAIL Innovation Lab
