LWLeykun WoldesemayatHire me on Upwork
AI Engineer · Addis Ababa · UTC+3

I build LLM-powered products. RAG systems, AI agents, and the full-stack apps around them.

Hire me on UpworkSee the Temar case study

Full-stack background in React and NestJS. Now focused on production systems built on large language models.

Portrait of Leykun Woldesemayat
4 yearsat ALX Africa
94learners graduated in one cohort
94%tech support CSAT
30,000+learners in my analytics pipelines
Featured work

Temar

AI-powered spaced repetition. It turns your notes into questions, grades your answers by meaning, and schedules each review for when you're about to forget.

Solo buildProprietarytemar.app ↗
temar.app
Temar dashboard showing due reviews, growth overview, study areas and the review schedule

Structured generation

Questions and rubrics arrive as typed schemas, validated before anyone sees them.

Semantic grading

Answers are graded against a rubric, then mapped to an FSRS rating.

Hybrid retrieval

pgvector plus Postgres full-text search, fused with Reciprocal Rank Fusion.

Cost control

Prompt caching and per-operation token costs against versioned pricing.

How it works

  1. 01You write notes
  2. 02 · LLMAI writes questions and rubrics
  3. 03You answer in your own words
  4. 04 · LLMAI grades by meaning
  5. 05 · FSRSThe next review is scheduled
Read the full case study

What I
build

Three kinds of projects I take on. Each one is already running in production in Temar.

More about me →

  1. LLM features in your existing app

    Generation, grading, summarization and chat, with the cost tracking to run them in production.

    Proof: Temar's tutor and semantic grading →

  2. RAG search over your documents

    Hybrid vector and keyword search on PostgreSQL and pgvector.

    Proof: Temar's hybrid search →

  3. Full-stack apps

    Next.js and NestJS, deployed with Docker and CI/CD.

    Proof: Temar's architecture →