Welcome

Backend Engineer building reliable systems for banking, payments & compliance

Hello, I'm Boyd Chanza

Software engineer · backend systems, API integrations & production AI

I build backend systems and API integrations for environments where failure isn't an option: banking infrastructure, national payment rails, and regulatory compliance systems that move real money. I also ship AI-assisted products end-to-end, from architecture to production.

Based in Lusaka, Zambia!

Let's Connect!

View case studies All projects

What I bring

Specialized expertise

A quick view of where I focus and the work behind it.

ISO 20022 Payment Migration

Hands-on experience moving live national payment infrastructure onto the ISO 20022 global messaging standard: a mandatory, deadline-driven shift most banks are still navigating. Not theoretical: implemented under regulatory pressure, on production systems, with zero missed settlement deadlines.

View case study

Core Banking System Integration & Certification

Led EFT/Cheque Truncation System rollouts and certification across 4 commercial banks, the kind of regulated, high-stakes integration work that most developers never touch and most agencies charge a premium for.

View case study

Secure AI Agent Architecture

Designed and built a production-style AI agent system (Athena) using Auth0 Token Vault: OAuth credentials and federated tokens never touch the application database. Most "AI agent" builds today skip this entirely; I built the trust boundary first.

View case study

Full-Stack Elixir/Phoenix Engineering

Deep, sustained experience in a concurrency-first stack most companies struggle to hire for: banking systems, AI-assisted products, and a multi-tenant VSDC integration gateway built specifically to exploit BEAM's isolated-process concurrency model.

View case study

Regulatory E-Invoicing Integration (ZRA Smart Invoice / VSDC)

Built a working ERP-to-VSDC integration for Zambia's mandatory e-invoicing system, the exact compliance gap most local businesses can't close in-house.

View case study

About Me

Welcome to my corner of the internet!

I'm a backend engineer with 6+ years across software development, systems support, and technical leadership in banking and fintech. I've led regulator-deadline migrations on live payment infrastructure, built multi-tenant API gateways with fault isolation, and shipped production systems that can't afford downtime. My core strength is designing reliable, API-driven backend architecture: PostgreSQL, MS SQL Server, Elixir/Phoenix, with growing depth in integrating AI into production systems the right way: secure credential handling, clear trust boundaries, human-in-the-loop where it matters. I care about systems that work in the real world, including low-resource settings, and I'm motivated by fintech and health-tech use cases where solid engineering actually changes how people access services.

Expertise

What I do

Backend & core systems

Design and maintain scalable backends, banking integrations (EFT, cheque truncation, ISO 20022), APIs, and data layers with PostgreSQL, MS SQL, and SQL-first thinking. Focus on uptime, reliability, and clear boundaries between services.

Systems leadership & delivery

Lead support and implementation teams, coordinate vendors, ship rollouts to financial institutions, and keep production systems above ~95% uptime, with disciplined documentation and stakeholder communication.

AI, data & applied analytics

Integrate AI into production systems the right way: secure credential handling, human-in-the-loop workflows, and clear trust boundaries (e.g. OAuth Token Vault, Vertex AI, Gemini). Build AI-assisted products end-to-end, not just prototypes.

Banking & fintech, in production

Case studies

Selected paid work where reliability and correctness were non-negotiable.

ISO 20022 EFT compliance upgrade for 4 commercial banks

Technical Lead · Probase · Lusaka

Situation: Zambia's national clearing house (ZECHL) mandated that all participant banks upgrade their EFT systems to the new ISO 20022 messaging standard, covering direct debits, direct credits, payment returns, and reversals. Each bank's core banking system had to produce and consume files in the new format to keep clearing payments through the national system.
Constraint: A regulator-driven deadline against live production payment infrastructure, executed as a single big-bang cutover: no phased rollout and no fallback to the old format. Any file-generation or validation error would hit real settlement traffic.
Action: Led the implementation for 2 of the 4 banks directly, and mentored and technically reviewed the team's work on the other 2, personally testing each system before go-live. As technical lead, I was accountable for correctness across all 4 implementations.
Result: All 4 banks went live on the mandated deadline. Post-cutover issues (~20% of expected, normal at this scale) were resolved within two weeks, stabilizing to a ~2% error margin. No issue involved delayed or failed fund transfers; errors were contained to non-critical validation and processing edge cases, not money movement.

Multi-tenant VSDC integration gateway (Elixir/BEAM)

ZRA Smart Invoice / VSDC · 2025

Situation: ZRA's Smart Invoice mandate requires every VAT-registered business to integrate with the government's VSDC, a locally-deployed service with a raw REST/JSON API. ZRA leaves each business to build and maintain its own integration, retry logic, and reliability layer; there is no shared, hosted gateway.
Constraint: A workable solution for multiple clients has to isolate each client's traffic and failures from every other client's: a genuine concurrency and fault-tolerance problem, not just an API wrapper.
Action: Designed and built a middleware API in Elixir/Phoenix, choosing the BEAM virtual machine deliberately: each request runs in its own lightweight, isolated process, so one client's failure or load spike cannot degrade another's. It is the same concurrency model that underlies systems like WhatsApp and has been benchmarked holding around 2 million simultaneous connections on a single machine.
Result: Built and functional; currently onboarding first clients.

Staying open for business while staying ZRA-compliant

Local-first Smart Invoice gateway · 2025 · invoice.digishift.io

Situation: Zambian businesses must fiscalize sales through ZRA's VSDC, but the regulator's API is low-level and often unreliable in production: device init, TPIN/branch/serial, classification codes, stock and purchase reporting, error codes, and retries. SMEs do not want to become integration engineers. They want to issue invoices and serve customers now, sync to ZRA when the device and network allow, and not lose a sale because VSDC is slow or down.
Constraint: Typical designs fail in production: syncing inside the user request makes cashiers wait; one tenant's VSDC outage can stall shared infrastructure; integrators re-implement the same VSDC flows in every project; duplicate submissions create reconciliation nightmares.
Action: Built a local-first invoicing and inventory product that also acts as a multi-tenant gateway to ZRA. Business actions land in PostgreSQL first (source of truth), then fiscalize as an explicit step: Sync to ZRA or POST /invoices/{id}/sync_zra, via durable Oban jobs with per-tenant VSDC config, circuit breaker, rate limit, retry, and duplicate (924) recovery. Shop owners use Phoenix LiveView; integrators use a documented REST API. Compliance covers more than sales: products, stock, purchases, credit notes, and imports.
Result: The sale can complete even when ZRA does not: the invoice exists locally and sync retries later. One tenant's failure does not block others. Integrators embed compliance once via REST instead of raw VSDC. Live at invoice.digishift.io.

Systems and API integrations for core banking platforms

Software Developer · Probase · Lusaka

Situation: Commercial banks running core banking platforms such as Finacle and Flexcube needed surrounding systems and API integrations to extend and automate their operations.
Constraint: Multiple institutions, high-stakes financial operations, and no room for downtime.
Action: Built and maintained systems that integrate with the banks' core banking platforms (Finacle, Flexcube) through API integrations, deployed applications, and ran incident reporting and enhancement cycles alongside production support.
Result: Delivered and supported across 4 commercial banks in Zambia.

EFT & cheque truncation reliability

Implementation & System Support Team Lead · Probase · Lusaka

Situation: Financial institutions depended on EFT and cheque truncation systems that had to stay available.
Constraint: Live financial infrastructure, vendor integrations, and strict uptime expectations.
Action: Led the implementation and support team, coordinated vendor integrations, and resolved defects and change requests.
Result: Sustained high uptime, including periods above ~95%, across client institutions.

Athena: production-style AI brand agent

Personal project · Elixir/Phoenix LiveView · 2026

Situation: Keeping a consistent brand presence across social platforms means constant content generation, scheduling, and publishing, work that is easy to automate badly and risky to hand fully to an AI.
Constraint: An AI agent that posts on your behalf needs access to third-party OAuth accounts (LinkedIn, Facebook, Google Calendar) without exposing those secrets or acting without oversight.
Action: Built an AI brand automation system in Elixir and Phoenix LiveView that generates, schedules, and publishes social content with human-in-the-loop approvals. Integrated Google Gemini, LinkedIn/Facebook posting, and Google Calendar-aware scheduling, and used Auth0 for AI agents with Token Vault so OAuth secrets never land in the app database. Kept a clear split between AI reasoning, identity, and execution, with PostgreSQL and Oban for background jobs.
Result: A working system where an AI agent drafts and schedules posts under human approval, with third-party credentials isolated from the application via Token Vault.

Niyah: AI insights from books and PDFs

Founder & lead engineer · Flutter + Elixir/Phoenix · 2025

Situation: Readers want the insights inside books and long PDFs without reading cover to cover, and want to try a product before committing to an account.
Constraint: Deliver AI understanding and audio affordably (unit economics matter) while removing signup friction for first-time users.
Action: Built a mobile app that turns books and PDFs into insights, summaries, and optional audio, using anonymous UUID onboarding, tiered pricing, and Google Vertex AI for understanding and text-to-speech. Flutter client with Elixir/Phoenix APIs, tuned for unit economics and a smooth path from anonymous use to a full account.
Result: A shipped app with a low-friction, anonymous-first flow, tiered pricing, and Vertex AI summaries and audio, taking users from a PDF to insights without an upfront signup.

Portfolio

Selected works

Smart Invoicing System

2025 · Zambia / ZRA integration

Local-first invoicing and inventory for Zambian SMEs, with a multi-tenant gateway to ZRA Smart Invoice / VSDC: records save locally, then fiscalize through durable background jobs. Phoenix LiveView for operators; documented REST API for integrators. Credit notes, product registration, stock, purchases, and QR receipts.

ElixirPhoenix LiveViewZRAVSDCFinTech

Visit site →

Niyah: AI insights from books & PDFs

Founder & lead engineer · 2025

Mobile app that turns books and PDFs into insights, summaries, and optional audio, with anonymous UUID onboarding, tiered pricing, and Google Vertex AI for understanding and TTS. Flutter client, Elixir/Phoenix APIs, tuned for unit economics and a smooth path from anonymous use to a full account.

FlutterElixirPhoenixVertex AIMobile

Visit site →

Athena: AI brand assistant

Personal project · 2026

Production-style AI brand automation in Elixir and Phoenix LiveView: generate, schedule, and publish social content with human-in-the-loop approvals. Integrates Google Gemini, LinkedIn/Facebook posting, Google Calendar-aware scheduling, and Auth0 for AI agents with Token Vault so OAuth secrets never land in the app DB. Uses PostgreSQL, Oban for jobs, and a clear split between AI reasoning, identity, and execution.

ElixirPhoenix LiveViewPostgreSQLObanGeminiAuth0

Visit site →

Stack & tools

Skills

Technologies and domains I work with day to day, aligned with my CV and recent projects.

Languages & frameworks

Elixir
Phoenix & LiveView
Flutter
SQL
R

Data & platforms

PostgreSQL
MS SQL Server
REST APIs
Oban

AI & integrations

OpenAI
Vertex AI
Gemini
Auth0
Token Vault

Banking & delivery

Core banking
EFT
Cheque truncation
Production support
Vendor coordination

Verified learning

Education & certifications

Data Analytics · ALX Africa
Data Analysis · Coursera
R Programming for Data Science · Udemy

Posts

Recent thoughts

2026 · Updates

Building Athena: agents, OAuth, and safe token handling

Notes on Auth0 Token Vault, private-key JWTs, and keeping AI agents on the right side of the trust boundary. More on LinkedIn.

Connect on LinkedIn →

2025 · Product

Shipping Niyah: from PDFs to insights without signup friction

Why anonymous-first onboarding, pricing tiers, and Vertex AI came together the way they did.

Connect on LinkedIn →

2024 · Data

Analytics, R, and real statements from mobile money

Lessons from classifying MTN and Airtel flows and turning them into dashboards users actually open.

Connect on LinkedIn →

Let's create something together

I'm open to conversations about backend engineering, AI-assisted products, fintech, and health tech, whether that's a project, a collaboration, or a quick question.

View case studies

LinkedIn Email 0973 767 852 WhatsApp