Every career has an interface.
The interesting part is underneath.
SIX CHAPTERS · 2021 → NOW · KEEP SCROLLING
Muthasir
Senior Software Engineer
DISTRIBUTED SYSTEMS, PLATFORM & EDGE ENGINEERING
I build across the complete device-to-cloud lifecycle — web products, IoT fleets, petabyte-scale data platforms, Kubernetes control planes, embedded Linux devices, and AI systems that act through real engineering tools.
BEFORE THE DESCENT — THE TOOLKIT
What I work in.
9 DISCIPLINES · SCROLL — EACH ONE JOINS THE LINE
Paired with AI, the impossible is just another build target.
HUMAN JUDGMENT × MACHINE LEVERAGE — NINE DISCIPLINES, ONE ENGINEER
CHAPTER 01 — 2021–2022
PRODUCT
ThunderCam Solutions · Web Developer (Freelance)
First, learn how complete products fit together.
I started by building complete web products end to end — not features, whole systems. E-commerce, social, and business-management platforms, each shipped with its payments, auth, operations, and admin surfaces working together.
WHAT I BUILT
- E-commerce platforms with guest carts, coupons, returns and invoicing
- OAuth and OTP verification flows
- Wallet and payment integrations
- Live chat with real-time messaging
- Business-management, school and travel platforms
- Admin applications alongside every customer surface
Complete web products and business systems — e-commerce, social applications, business management, school systems, travel.
PAYMENTS · OAUTH · OTP AUTH · WALLETS
COUPONS · INVOICING · GUEST CARTS · RETURNS
LIVE CHAT · ADMIN APPLICATIONS
Light moves between the panes: a guest cart flowing into payment, an OTP round-trip, a return travelling back up.
By the bottom of the stratum the scattered surfaces have organised into one coherent system — the first hint that everything has an underneath.
CROSSING 01→02
THE SCREEN BREAKS
The last pane fills the viewport — a fuel-delivery order screen. The camera passes through it, the wireframe dissolves into heat haze, and you emerge in blinding daylight.
CHAPTER 02 — 2022–2024
FIELD SYSTEMS
Romulus Oil & Gas · Full Stack Engineer
In 2022, the software left the building.
I planned, built, deployed and maintained fuel-management and diesel-delivery platforms — web and mobile applications connected to real hardware in the field. This is where my software first had to talk to microcontrollers, tanks and trucks, not just browsers.
WHAT I BUILT
- IoT integrations: live fuel-level monitoring via field microcontrollers
- Location-based asset tracking and real-time delivery operations
- Billing, invoicing and online ordering
- Live-chat support with real-time functionality
- Telemetry pipeline: field hardware → cloud → backend → web/mobile
- Cloud right-sizing, caching and asset compression on existing apps
Fuel-management and diesel-delivery systems: planned, built, deployed, maintained.
THE ONLY DAYLIGHT ON THIS SITE — WHERE SOFTWARE MET THE PHYSICAL WORLD
Field hardware → microcontrollers → telemetry → cloud services → backend systems → web and mobile applications.
IoT INTEGRATIONS · FUEL-LEVEL MONITORING
ASSET TRACKING · LOCATION SYSTEMS · TELEMETRY
Real-time delivery operations — online ordering, billing, invoicing, live communication — running against moving trucks in the heat.
Optimised the existing applications. The plate reads:
> 3× PERFORMANCE IMPROVEMENT
≈ 50% SERVER-COST REDUCTION
CROSSING 02→03
THE BORE
The camera dives into a tank’s sensor well — daylight shrinks to a circle above — and the shaft floor turns out to be a surface. You break through it, upside down, into an ocean.
CHAPTER 03 — 2024
DATA PLATFORM
Aggregate Intelligence · Full Stack Developer
Then the data got heavy.
I engineered data-intensive systems over petabyte-scale travel datasets inside an enterprise ecosystem serving global online-travel, hospitality and aviation products. Data at that scale is a fundamentally different discipline from application development — storage layout and orchestration decisions have real cost and latency consequences.
WHAT I BUILT
- Delta Lake table and storage configurations for petabyte-scale distributed datasets
- Spark-based distributed processing on Google Cloud Dataproc
- Workflow orchestration with Cloud Composer / Airflow
- Backend services and interfaces with FastAPI and React
- CI/CD automation that shortened release cycles
- Code reviews and mentoring across the team
THE DATA ECOSYSTEM SERVED GLOBAL TRAVEL LEADERS
Petabyte-scale datasets. A data ecosystem supporting global travel, hospitality, flight and aviation use cases.
10¹⁵ BYTES — THE WIDEST SHOT ON THE SITE
A Spark job crosses the ocean as a wavefront — partitioning, shuffling, re-collecting the field around you.
APACHE SPARK · DATAPROC · COMPOSER / AIRFLOW
BIGQUERY · DELTA LAKE
Near the floor, the currents crystallise into ordered slabs — Delta Lake table and storage configurations for distributed systems.
CROSSING 03→04
THE INTAKE
Near the ocean floor the currents all bend toward one point — a machined circular intake grate, lit from below. The particles, and you, are drawn through it.
CHAPTER 04 — 2024 — PRESENT
PLATFORM & CONTROL PLANE
Nuventure Connect · Software Engineer
Machines that run machines.
I work across the architecture and implementation of NuWave — a Nuventure product, the multi-service platform behind its pool-automation and IoT lines — distributed backend services, agent infrastructure, device-facing systems and control-plane capabilities. The defining work: software that uses Kubernetes as a dynamic workload-execution platform — programmatically creating, running and observing containerized agent workloads — not just deploying to it.
WHAT I BUILT
- Agent Management & Agent Launcher: on-demand pod creation via the Kubernetes API, execution lifecycle, runtime state surfaced to the control plane
- Containerized microservices and platform APIs with service-to-service communication on Kubernetes
- Device anomaly detection over telemetry, runtime statistics and operational signals
- Data lakehouse: Spark, Iceberg, MinIO, Trino, Superset — open table formats, object-storage-backed analytics, separation of storage and compute
- Production platform infrastructure: ingress-nginx, cert-manager/TLS, Cloudflare DNS, GitLab CI/CD, container registries
- Production debugging across ingress, upstream routing, DNS, certificates, container networking and registry auth
- Tool-connected AI workflows and cross-platform Flutter applications
THE PHYSICAL PRODUCT DOMAIN — POOL AUTOMATION & IoT
NuWave: a multi-service platform — microservices, infrastructure, agent systems, device-facing services, operational and AI capabilities.
WORKED ACROSS ARCHITECTURE AND IMPLEMENTATION
OWNED KEY SYSTEMS AND COMPONENTS
The Agent Launcher: scroll drives the loop. Control plane → Agent Management → Kubernetes API → pods materialise → execution state flows back.
NOT DEPLOYING SOFTWARE TO KUBERNETES —
WRITING SOFTWARE THAT COMMANDS IT.
A wing of instruments watches the fleet: one telemetry stream drifts out of family, and a crimson ring closes around it.
DEVICE ANOMALY DETECTION — TELEMETRY · RUNTIME STATS · OPERATIONAL SIGNALS
CROSSING 04→05
POTTING
The camera pushes through a running pod into the node beneath it, and the world sets around you like black epoxy. A faint warm glow appears below: copper.
CHAPTER 05 — NUVENTURE — PROJECT
FLEETMAN
Nuventure Connect · Software Engineer
The bottom of the stack has a temperature.
FleetMan is an edge fleet-management platform I work on at Nuventure: customized BalenaOS device environments, provisioning and identity, a device-side runtime, remote container-workload lifecycle, and cloud-side fleet control. It runs on constrained ARM hardware, where I debugged container execution down to CNI networking and IPv4 forwarding.
WHAT I BUILT
- Device-side runtime/supervisor in Python + aiohttp: API-driven workload updates, container deployment, restarts, status, remote logs, device statistics
- Container execution on ARM Linux with containerd and nerdctl — resolving CNI, snapshotter, image-layer and registry-auth issues
- Customized BalenaOS environments, device provisioning, fleet registration and identity
- Embedded Linux integration on Variscite DART-6UL / NXP i.MX6UL(L): Device Tree, sysfs, GPIO, UART, SPI, 1-Wire, Modbus
- Industrial sensor integration (PT100, DS18B20) down to register level
- Cloud-side fleet control with remote operations and observability
FleetMan — an IoT and edge fleet-management platform I work on at Nuventure: customised BalenaOS environments, provisioning, identity, lifecycle, remote deployment.
A lightweight device-side runtime and supervisor in Python + aiohttp: workload updates, container deployment, restarts, service status, remote logs, device statistics.
containerd · nerdctl · CNI · SNAPSHOTTERS
ARM IMAGES · GITLAB REGISTRY · IPv4 FORWARDING
At the board’s edge, a single wire leaves the copper world into darkness — a DS18B20 on a 1-Wire bus, drawn oscilloscope-grade.
UART · SPI · GPIO · 1-WIRE · MODBUS · sysfs · DEVICE TREE
FROM PETABYTES TO PULL-UPS.
Pull back: the one board is one lime point in a constellation — provisioning, registration, identity, lifecycle, remote operations, cloud-side fleet control.
CROSSING 05→06
IGNITION
At the deepest point, a probe tip touches the copper trace. A pulse of lime enters the trace — and instead of dissipating, it turns upward. The depth gauge ticks up for the first time.
CHAPTER 06 — NOW
THE CURRENT REVERSES
Software that acts through other software.
Much of my range comes from independently investigating systems because I wanted to understand them. This work starts as technical exploration and becomes working software — and lately it converges on one theme: AI that participates in real engineering workflows instead of just chatting.
WHAT I BUILT
- AI-native website generation platform — built before AI website builders were a mainstream category (Git history as evidence): model-generated structured content, React template composition, runtime JSX rendering
- Tool-connected AI workflows: LLM systems driving Blender and KiCad for model-driven action execution
- ONNX model inference and browser deployment research (BiRefNet, InSPyReNet)
- AIoT operational intelligence: AI-generated telemetry summaries and device anomaly reasoning
Passing bedrock: the current threads into device telemetry — AI-generated summaries, operational reasoning. Dashboards that interpret, not just display.
Passing the engine room: AI-enabled platform systems and containerised agent workloads — the Launcher and the current shake hands.
Passing the ocean: inference placement — cloud, local, browser, edge — the question of where intelligence should physically run.
Passing the glass: the current builds a website pane by itself — AI-generated structured content, template systems, runtime-rendered JSX. Built before the category was mainstream.
GIT HISTORY AS EVIDENCE
Above the original interface, new territory: LLM → tool interface → domain application → action → result. Blender builds a mesh. KiCad routes a schematic.
INTEREST: AI THAT PARTICIPATES IN ENGINEERING WORKFLOWS — NOT CHAT.
Every layer on this descent is still running.
Nothing was left behind — it compounded.
MUTHASIR
Senior Software Engineer — Distributed Systems, Platform & Edge
GIVEN A LAYER, HE SHIPS THE LAYER. THEN HE GOES BELOW IT.