Wajed Afaneh

Selected work

Four pieces of work I would do again. Each is the problem as it looked at the start, what I actually did, and what changed.

Turning graduates into engineers

ASAL Technologies, 2017 onwards

Palestine produces far more computer-science graduates than its market can absorb, and employers keep saying the graduates are not ready. We needed senior engineers we could not hire, and the people who could become them were sitting outside the industry.

I designed a train-to-hire programme: a curriculum I wrote and taught, assessment at each stage, placement on live client teams with a named mentor, and structured feedback every quarter. Part of it ran under the World Bank-funded Train-to-Hire initiative. Alongside it I set up dual-studies placements with An-Najah, Birzeit and Al-Quds universities.

What changed

  • More than 30 engineers placed.
  • About 85% still with the company two years later.
  • Roughly 40% of the current senior bench started in the programme, and several now lead teams.
  • 50+ university students hosted; more than ten now work on my teams or in senior roles.

Moving a monolith without stopping it

ASAL Technologies, enterprise customers on AWS and Azure

A monolithic system was slowing every release. Deployments took weeks to prepare and incidents were frequent, but the business could not accept downtime while we changed it.

I designed the service boundaries, data ownership, API contracts and deployment topology with the senior engineers, and ran the migration in slices. Service-level objectives were agreed before we started, alerting was built on what users feel, and every cutover had a rehearsed rollback.

What changed

  • Zero-downtime cutover.
  • Deployment lead time from about three weeks to under two days.
  • Production incidents down by around 60%.
  • Releases moved from monthly to several a week, with rollback under ten minutes.

Putting AI into production, not into a demo

ASAL Technologies, a consumer marketplace

The marketplace wanted AI features, and its listing classification ran through a manual review queue with a 48-hour turnaround that could not keep up with growth.

I led OpenAI integrations across three core flows, then fine-tuned a LLaMA-based model on the client's own review history so listing data never left their environment. We judged the model against the human reviewers, not a benchmark, and kept it under drift monitoring.

What changed

  • About 25% higher engagement on the three AI-assisted flows.
  • About 70% of listing classification automated.
  • Manual review turnaround from 48 hours to under four.

Building an engineering team from nothing

OptionRoom, 2020 – 2022, as CTO

OptionRoom had a product idea, investors and a community, and no engineering team.

I hired the team, set the two-year roadmap with the CEO and investors, built the platform on AWS and GCP with containers and automated delivery, and rewrote the early prototype as production services. We open-sourced the code and I explained architecture and trade-offs to the community in public AMA sessions.

What changed

  • A team of five shipping an oracle platform with about fifty data feeds across four blockchains.
  • Release cycle about five times faster than the prototype.
  • Bad external data flagged in under two seconds before it reached the chain.
  • Engineering support through a seed round of about $400K and token sales of about $600K.