Selected work

Systems that make the operation clearer, faster, and harder to break.

These are examples of the problems I take on—not a catalog of every tool I have used. The important part is what changed for the people running the business.

Operational visibility at scale

From scattered information to a working picture of statewide operations.

A growing network of remote sites had no unified business intelligence for site health, assets, personnel, repairs, or recurring failures. I combined Jira workflows, InvenTree inventory, PagerDuty, Uptime Kuma, AI-assisted reporting, cloud services, and underused company servers into an operational layer the team could use.

Created visibility across approximately 1,260 devices at 21 sites and counting.
Enabled the team to identify and address problems before customers reported them.
Improved the evidence available for supplier, customer, and third-party audits.
Translated alerts and technical data into clearer reporting for executives.

21+ sitesstatewide footprint

~1,260 devicesmixed environment

630–840Linux systems

Continuity and repair

Repair time stopped being the same thing as customer downtime.

The inherited system design lacked meaningful redundancy, business continuity, and disaster recovery. A repair involving an overseas team could leave one site waiting for weeks. I introduced redundancy, proposed out-of-band management, established better repair tracking, and helped move the workshop toward centralized system images and hot-swappable replacements.

Reduced comparable repair turnaround from weeks to approximately 24 hours.
Used hot-swapping to keep the customer from experiencing the workshop repair interval.
Turned repair history into evidence for broader system upgrades.
Reduced the need to send someone to a remote site simply to restore access.

Cloud efficiency and integration

Technical architecture decisions made with business runway in mind.

As a Software Engineer in R&D, I worked across partner integrations, modular deployments, cloud infrastructure, data systems, and remote RF and ALPR hardware. I also contributed to AWS cost reductions that maintained prototype functionality while helping extend the company’s operating runway.

Led technical collaboration across partners including Intel, Snowflake, MongoDB, ArangoDB, AWS, DigitalOcean, and Neo4j.
Designed modular systems for remote data collection, transformation, and distribution.
Built Python services, device automation, and data-processing integrations.
Contributed cloud cost controls that helped extend runway by nine months.

Cloud + on-premdistributed architecture

Pythonautomation and services

RF + ALPRcustom capture systems

Automation at the bench

A repeatable setup process replaced slow, manual onboarding.

The Speedy Integration Tool used PowerShell to standardize new-computer setup and reduce technician effort. It handled naming, software installation and removal, Windows and OEM updates, and desktop shortcuts while making the process easier to teach and troubleshoot.

Automated common setup and update tasks.
Reduced avoidable variation between technicians.
Made onboarding faster to execute and easier to train.
Kept the workflow documented and maintainable.

Technical projects

Learning by building the whole thing.

Personal and academic projects are where I test ideas across software, hardware, data, and infrastructure.

Reinforcement learning

Self-contained Solana trading system

A Python trading system using reinforcement learning, Docker Compose services, MongoDB time-series data, historical backtesting, and custom indicators.

Remote data capture

Small-scale photo pipeline

An ESP32 camera pipeline that captures vehicles remotely, calls a Flask API, and routes images to a home system for vehicle recognition and low-cost license plate recognition.

Have a similar problem?

Start with the part that keeps failing.

You do not need a polished scope. A clear description of what is not working is enough to begin.

Email Alvan