Senior Engineer · Vancouver, Canada · Open to Opportunities

@madmmas

◆ Witness Me ◆

Senior Software & Data Engineer

$ ./madmmas --spec distributed-systems · cloud-native · AI-pipelines ▋

Senior software & data engineer with 15+ years building distributed systems at scale — from fintech payment infrastructure at American Express UK to AI-native pipelines and Kubernetes operators. I care about how systems are built as much as what they do.

Distributed systems & event-driven architecture — Kafka, Go, Java
Cloud-native & platform engineering — AWS (6×), Kubernetes, Terraform
AI/ML systems — RAG pipelines, LLM integration, MLOps
Open source — temflowral, KubeAgent, Queriva, AIPlane, News Radar

15+ yrs exp

6× AWS certs

8 OSS repos

456+ LeetCode

// GitHub Explore Work Blog ↗

Projects · Open Source & Production

What I've built

Projects that show
how I think.

Live ProductAI / MLDevOps · MLOps

AnalyseThisWC26 — FIFA WC2026 Analytics

Live analytics & AI match-prediction platform for FIFA World Cup 2026 by NeuNov Technologies. ESPN data ETL, Elo / Dixon-Coles / XGBoost / LightGBM blended prediction, Next.js interactive dashboards.

My contribution — DevOps & MLOps

Full AWS infra: ECS Fargate, S3, DynamoDB, Lambda — provisioned with Terraform
ETL pipeline: ESPN scraping → transform → DynamoDB / S3, triggered by Lambda
GitHub Actions CI/CD — model retraining, staged deployment: Jupyter artefacts → staging → production
Infrastructure coordination for Python prediction models and Next.js frontend

AWS ECSLambdaS3DynamoDBTerraformGitHub ActionsPythonXGBoostLightGBMNext.js

atwc26.com — live site →

Infra

AWS + Terraform

ECS · Lambda · S3 · DynamoDB

Data

ETL Pipeline

ESPN scraping → Lambda → DynamoDB

MLOps

CI/CD + Promotion

GitHub Actions · Jupyter artefacts

Models

Blended Prediction

Elo · Dixon-Coles · XGBoost · LightGBM

Live MockSaaS PlatformProduct Lead

Swavu — Rwanda Mobile Commerce & Admin Platform

Digital transformation for Swavu Ltd, Rwanda-based social enterprise. Led product design across: consumer mobile app (QR verification, loyalty, MoMo), admin console, and marketing website — with full architecture documentation.

My contribution — Product & Platform Architecture

Mobile app prototype: QR verification (Swavu Promise), loyalty engine, shop locator, MoMo-integrated returns
Admin console architecture: catalogue, supplier management, order fulfilment, analytics dashboards
Formal artefacts: RPI doc, system architecture, API contracts, Go backend design
Rebuilt marketing website from concept to deployed mock; sole technical and product lead

GoNext.jsReact NativeAWSVercelQR VerificationLoyalty EngineMoMo

Website → Mobile → Admin →

Surface 1

Mobile App

React Native · QR · Loyalty · MoMo

Surface 2

Admin Console

Next.js · Catalogue · Orders

Surface 3

Website

Next.js · Vercel

Backend

Go API Design

Architecture + contracts

OSSActive

temflowral

Visual durable-workflow editor connecting React Flow with Temporal's execution engine. Drag nodes to define long-running process logic — Temporal guarantees fault-tolerant, exactly-once execution behind the scenes. Ships as open source with full ADR documentation and a 21-day execution plan.

React FlowTypeScriptTemporalGoGitHub Actions

github.com/madmmas →

Type

Open Source · MIT

Core idea

Visual → Durable

React Flow nodes → Temporal workflows

Status

Active development

21-commit execution plan

AI / MLOSS

KubeAgent Control Plane

Go multi-module Kubernetes operator with three CRDs (Agent / Finding / RemediationPlan) and AWS Bedrock integration for AI-driven cluster remediation. Designed as a 21-commit progressive learning guide for engineers building Kubebuilder operators — from scaffold to production-grade reconciler loops.

GoKubebuilderAWS BedrockCRDsKubernetes

github.com/madmmas →

Type

Open Source

K8s Operator

Core idea

AI meets K8s

Bedrock LLM → cluster remediation

CRDs

Agent · Finding

RemediationPlan

AI / MLOSS

News Radar

Multilingual news aggregator for Bangladeshi sources. LaBSE embeddings generate semantic vectors; Qdrant stores and retrieves them; a five-service Kafka pipeline fans news events across processors; and a cascaded LLM-judge (Mistral → Claude Haiku → Claude Sonnet) makes final deduplication decisions at calibrated cosine thresholds. Validated across 800+ article pairs.

PythonLaBSEQdrantKafkaAWS BedrockMistral

github.com/madmmas →

Type

OSS · Research

Bengali / multilingual NLP

Core idea

Vector clustering

LaBSE + LLM judge cascade

Thresholds

0.60 reject

0.80 auto-accept

SaaSPrivate

Ackloop

Operational voice automation SaaS for NOC teams, hospital on-call, and logistics dispatchers. Full PostgreSQL event-sourced data model, 18-endpoint API contract, and Kafka topic contracts for real-time alerting and voice acknowledgement workflows. Powered by Deepgram Nova 3, ElevenLabs, Temporal Cloud, and Upstash Redis.

GoPostgreSQLKafkaReactTemporal CloudDeepgramElevenLabs

Private repository

Type

Commercial SaaS

NOC / Ops / Hospital

Core idea

Voice + Kafka

Real-time alert acknowledgement

API

18 endpoints

Event-sourced PostgreSQL

OSS Active Micro-Frontend

AIPlane — AI Management Control Plane

Self-hostable micro-frontend monorepo for LLM-powered application management. A Vite Module Federation host/remote architecture loads five independent federated apps into a single dashboard — each deployable and developable in isolation. Built on pnpm workspaces + Turborepo for build-order-aware caching.

Architecture highlights

Module Federation — Dashboard (host) dynamically loads Prompt Manager, Guardrail, User Manager, Usages Data as independent remotes at runtime
Turborepo — build pipeline respects dependency order; incremental caching cuts rebuild times across 5 apps
Each remote runs independently on its own port — zero coupling between feature teams at the module level
Full SPEC.md + branding — logo assets, design system, and product specification completed before first line of feature code

React 18TypeScript ViteModule Federation Turborepopnpm workspaces Tailwind CSSshadcn/ui

github.com/madmmas/aiplane →

Architecture

Module Federation

Vite Plugin Federation

Apps

5 federated remotes

Dashboard · Prompt Mgr · Guardrail · User Mgr · Usages

Monorepo

pnpm + Turborepo

Build-order caching

Status

Scaffold complete

15 tracked issues

OSS AI / RAG Local-First

Queriva — Ask Your Data Anything

Self-hosted, local-first LLM semantic search engine — zero runtime calls to external APIs. Accepts natural language queries, embeds them with a local model via FastAPI sidecar, searches a Qdrant vector collection, and optionally synthesizes answers with a local LLM via Ollama. Embeddable as a Module Federation widget into host apps like AIPlane and News Radar.

Architecture highlights

Dual mode — fast ranked vector search OR full RAG synthesis with Ollama; toggled per query
FastAPI embed sidecar — embedding generation isolated from the Java API; swap models without redeploying the gateway
Spring Boot 3 / Java 21 search + ingest gateway with Testcontainers integration tests running against real Qdrant
Embeddable MFE widget — consumable via Module Federation in AIPlane, News Radar, or any Vite host
Full SPEC.md + 31-issue implementation backlog written before scaffold; architecture-first development

Java 21Spring Boot 3 FastAPIPython QdrantOllama React 18Vite MFE TestcontainersDocker

github.com/madmmas/queriva →

Tagline

Ask your data anything

Local-first · no external APIs

Qdrant vectors

Natural language → ranked results

RAG

Ollama synthesis

Grounded LLM answers

Status

Monorepo scaffolded

31-issue backlog · SPEC.md done

Armoury · Specialisms

Technical depth

Where I operate
at my best.

Not a generalist who touches everything — an engineer with specific depth in distributed systems, cloud-native platforms, and AI/ML pipelines. These are the problem spaces I find genuinely interesting.

Distributed Systems

Microservices that hold up under real load. API contract discipline, structured error handling, distributed tracing, and SLO-driven observability — not bolted on later.

GoJavaSpring BootgRPCREST

Event-Driven & Streaming

Kafka at production scale — topic design, consumer group management, Schema Registry, exactly-once semantics, and the CQRS / Event Sourcing patterns that make state auditable.

KafkaAvroCQRSEvent Sourcing

Cloud-Native & Platform Engineering

AWS infra designed for teams, not tutorials. Kubernetes operators, Terraform modules, GitOps pipelines, and MLOps workflows that treat model deployment like software deployment.

AWSKubernetesTerraformGitHub Actions

AI / ML Systems Engineering

The plumbing underneath AI products — RAG pipelines, vector stores, embedding services, LLM gateway design, evaluation harnesses. The parts that make models reliable in production.

AWS BedrockLangChainpgvectorRAG

Data Engineering

Schema design that survives product changes. ETL/ELT pipelines with proper lineage, idempotent processing, and failure recovery — not just scripts that work once.

PythonPostgreSQLDynamoDBRedshift

Technical Writing & OSS

Engineering blog covering Go concurrency, Kubernetes operators, AWS architecture, and Java internals. Five OSS projects with full ADRs, SPEC.md, and structured commit histories.

ADRSecurityPerformanceScalability

War Code · Engineering Values

How I operate

What I believe about
building software.

Production is the spec

15 years of systems that had to work — AmEx UK card transactions, fibre network provisioning, multilingual NLP pipelines. I design for the failure modes, not the happy path.

Architecture decisions are written down

ADRs, SPEC.md, API contracts — written before the first line of implementation. Future engineers should understand why, not just what.

Distributed team native

UK, Bangladesh, Canada — three continents of async collaboration. Strong written communication is a non-negotiable engineering skill, not a soft skill.

Complexity budget matters

The best system is the one your team can reason about at 2am. I default to boring technology for solved problems and save the interesting choices for unsolved ones.

Credentials backed by practice

6× AWS certified + CKAD — each certification reflects systems I've actually designed and operated. The LeetCode habit (456+ problems) is about keeping the fundamentals sharp.

Teaching is part of the job

Engineering blog, five OSS projects with structured commit guides, 50-problem Java concurrency repo. If I can't explain it clearly, I don't understand it well enough.

War Record · Experience

Career record

15 years of systems
shipped to production.

2022 — Jan 2025

LTI Mindtree

American Express UK · London

Senior Software Engineer

Led event-driven microservices processing millions of daily card transactions — Kafka, Java Spring Boot, AWS.
Cloud migration of legacy systems via Terraform IaC; significant infrastructure cost reduction.
Established contract-first API design, ADR culture, and SLO dashboards across the engineering team.

2019 — 2022

Truespeed Communications

Bristol, United Kingdom

Senior Software Engineer

Built the fibre broadband provisioning and CRM platform from scratch — Java, Spring Boot, PostgreSQL.
Real-time network telemetry pipeline for field operations; integrated OSS / BSS systems.

2009 — 2019

Nitto Digital · Kiteplex IT

Dhaka, Bangladesh

Technical Lead → Senior Engineer

Led full-stack teams; introduced microservices architecture and CI/CD across product lines.
Mentored engineers; drove adoption of automated testing and clean code practices.

Combat Medals · Certifications

Verified credentials

Earned, not inherited.

AWS Data Engineering Associate

DEA-C01

AWS Generative AI Professional

ACP-GEN

AWS Solutions Architect Associate

SAA-C03

AWS Developer Associate

DVA-C02

AWS AI Practitioner

AIF-C01

CKAD — Kubernetes Application Developer

CNCF · Linux Foundation

The Citadel · Get in Touch

Let's build
something worth it.

I'm open to senior engineering roles, staff-level positions, and interesting technical problems. If you're building something with genuine technical ambition — reach out. I respond to every message.

// GitHub

GitHub

github.com/madmmas · OSS projects + activity

→

linkedin.com/in/madmmas

→

Engineering Blog

madmmasblog.vercel.app

→

Preferred for role enquiries

→

Download CV