Skip to content
BEN JAHSON.

About

Developer by craft. Evaluator by discipline.

A closer look at how I work across web development and AI evaluation.

01 — About

Useful products, carefully made.

BJT

I build modern, responsive digital experiences that help businesses establish their online presence and operate more effectively. I care about products that are useful, accessible, and quietly well designed — the kind people can use without thinking about it.

Alongside software development, I contribute to AI training and evaluation workflows, helping improve the quality, accuracy, and usefulness of AI-generated outputs. Reviewing how models respond has sharpened how I think about clarity, correctness, and the people on the other side of the screen.

The two disciplines inform each other: building teaches me what good output looks like in practice, and evaluating teaches me to question it carefully.

  • 01Practical website development
  • 02Frontend and backend development
  • 03Responsive user interfaces
  • 04Business-focused digital solutions
  • 05AI model training and evaluation
  • 06Quality assessment of AI-generated content

02 — AI Training & Evaluation

Working With AI. Improving What Comes Next.

Through project-based work on AI training and evaluation platforms, including Outlier and Handshake, I review and assess model outputs against detailed guidelines. The work is careful and methodical: reading closely, judging fairly, and writing feedback that helps models become more accurate and more useful.

Platform-based project work · Outlier · Handshake

  1. 01

    AI response evaluation

    Reviewing language model responses for accuracy, helpfulness, tone, and completeness.

  2. 02

    Model output quality assessment

    Rating outputs against structured rubrics and explaining the reasoning behind each judgement.

  3. 03

    Instruction-following assessment

    Checking whether a response actually does what the prompt asked — including constraints and format.

  4. 04

    Text-to-image output evaluation

    Assessing how faithfully generated images reflect the written prompt.

  5. 05

    Reference-image consistency

    Comparing generated images against reference inputs for identity, style, and detail consistency.

  6. 06

    Visual quality & alignment

    Flagging artefacts, composition issues, and outputs that conflict with responsible AI guidelines.

Start a conversation