Liben Adugna
Software engineer working remotely, with full overlap with European working hours. I build ML-backed products with Python, Django and Flutter, and I author the evaluation benchmarks used to test AI coding agents.
libenadugna285@gmail.com · GitHub · LinkedIn · Resume (PDF)
Selected work
- Smart Gebere: Flutter app that helps smallholder farmers diagnose crop disease with on-device ML, built for low-connectivity use.
- Benchmark and evaluation design for AI coding agents: Terminal-Bench 2.0 and 3.0 tasks, SWE-bench style tasks, Docker environments with hidden tests, MongoDB and PostgreSQL backed tasks.
- Behavioral user segmentation on roughly 450,000 interaction records (Python, scikit-learn, K-Means, PCA, silhouette validation).
- Transaction anomaly detection on roughly 110,000 records (Isolation Forest, Random Forest, threshold tuning for precision).
- Full-stack e-commerce: Django REST backend with Stripe payments and a Flutter client.
- This site: React, TypeScript, Tailwind, with a Gemini-powered assistant behind a serverless endpoint.
Experience
- Freelance machine learning engineer, remote (2022 to present)
- AI benchmark task author: Terminal-Bench 2.0 and 3.0 at Revelo, Terminal-Bench 2 (Project Claw) at AfterQuery, SWE-bench style tasks at Turing (2024 to present)
- AI systems evaluator and data architecture contributor, Revelo (2025 to 2026)
- AI model evaluator and training specialist, Afriwork / Turing (2024 to 2025)
- Full-stack software engineer, Faris Technology (2023 to 2025)
- Senior technical mentor, machine learning, CSEC university computing club (2022 to 2023)
Education
- B.Sc. Software Engineering (2022)
- Africa to Silicon Valley (A2SV) algorithms program, 400+ problems solved