Open to AI / ML roles & research collaborations

Hi, I'm

Yabsera Haile Yemanberhan

</>

I build applied AI, LLM systems, and AI agents that solve real-world problems.

View my work Download résumé

Kigali, Rwanda

Scroll

01 — About

Engineer at the intersection of AI & real-world impact

AI Engineer, Machine Learning Engineer, and Software Engineer with 5+ years of experience building web applications, machine learning systems, and scalable software. My work spans applied AI, LLM systems, AI agents, and deploying ML models for real-world use — across healthcare AI, multilingual NLP, full-stack development, and cloud-native applications. I thrive in collaborative teams and love tackling hard engineering problems end to end.

Applied AI & LLM systemsMultilingual NLP researchML model deployment

Highlights & achievements

Primary author on a research paper accepted at COLM 2026
AWS AI Practitioner & Cloud Essentials Certified
Top 1.2% of 5M+ LeetCode users — 600+ problems solved
Salutatorian, Software Engineering Department — AASTU
Completed the 15-month ALX & ExploreAI Data Science Program

Years building software

3.94 / 4.0

GPA at Carnegie Mellon

Years of AI Research and Engineering

Peer-reviewed Research Paper Published

02 — Toolkit

The stack I build & research with

From low-level CUDA and deep-learning research to production full-stack systems.

PythonPyTorchTensorFlowCUDALLM SystemsTransformersScikit-LearnTypeScriptNext.jsReactNode.jsExpressAWSDockerPostgreSQLMongoDBReact NativeAudio MLPythonPyTorchTensorFlowCUDALLM SystemsTransformersScikit-LearnTypeScriptNext.jsReactNode.jsExpressAWSDockerPostgreSQLMongoDBReact NativeAudio ML

Languages

PythonCUDATypeScriptJavaScriptJavaSQLDart / Flutter

Machine Learning

PyTorchTensorFlowScikit-LearnTransformersAudio ML

Cloud & DevOps

AWSDocker

Backend

Node.jsExpress.jsREST APIsSequelizeMongoose

Databases

PostgreSQLMySQLMongoDB

Frontend

ReactNext.jsReduxTailwind CSSMaterial UIReact Native

03 — Education

Where I trained

Graduate AI research at Carnegie Mellon, built on a software-engineering foundation.

CMU

GPA 3.94 / 4.0

Carnegie Mellon University

Master of Science in Engineering Artificial Intelligence

Aug 2024 – May 2026Pittsburgh, PA, USA

Graduate research in multilingual NLP and machine-translation detection.
Concentration in LLM systems, AI agents, and applied deep learning.

Relevant coursework

Introduction to Deep LearningLLM SystemsAI System DesignMachine Learning for Signal ProcessingApplied Stochastic ProcessesData Intensive Applications of Machine LearningMachine Learning for EngineersAI Agents for Engineering

Degree Certificate Transcript

AASTU

GPA 3.85 / 4.0

Addis Ababa Science and Technology University

Bachelor of Science in Software Engineering

Oct 2019 – Jul 2024Addis Ababa, Ethiopia

Graduated as Salutatorian of the Software Engineering Department.
Strong foundation in systems, security, and distributed computing.

Relevant coursework

Data Structures and AlgorithmsOperating SystemsDatabase DesignComputer SecurityDistributed Systems

Degree Certificate Transcript

04 — Experience

Where I've worked

From graduate AI research to product and full-stack engineering.

Carnegie Mellon University

Current

Research Associate — Languages & Spatial Technologies Lab

Visit website

Sep 2025 – PresentKigali, Rwanda

Lead research on machine-translated vs. human-translated text detection across English, Spanish, Swahili, and Afrikaans.
Built and evaluated transformer classifiers including mDeBERTa and XLM-R across sentence-level and long-context settings.
Supported dataset construction with 120,000+ labeled examples across multiple domains and translation systems.
Evaluated GPT-5, Gemini 2.5 Pro, and Claude Opus 4 for multilingual machine-translation detection.
Co-authored a research paper currently under review at COLM 2026.
Promoted from Research Assistant to Research Associate.

NLPLLMsTransformersResearch

Carnegie Mellon University

Teaching Assistant — Introduction to Deep Learning

Visit website

Jan 2026 – May 2026Kigali, Rwanda

Support delivery of graduate-level deep learning coursework covering CNNs, Transformers, optimization, and modern architectures.
Guide students through office hours, assignment help, and project reviews.
Evaluate assignments and give feedback on model implementation and experimentation.

Deep LearningTeachingCNNsTransformers

Carnegie Mellon University

Research Assistant Intern

Visit website

Jun 2025 – Sep 2025Kigali, Rwanda

Collaborated with engineers and medical doctors from CMU and Makerere University on a machine-learning tuberculosis screening system using cough audio.
Preprocessed and analyzed 92+ hours of cough audio using MFCC and spectrogram-based feature extraction.
Built and evaluated Logistic Regression, CNN, and Audio Spectrogram Transformer models for TB detection.
Optimized lightweight deep learning models for offline mobile deployment in low-connectivity environments.
Contributed to an AI-powered mobile application for tuberculosis screening.

Healthcare AIAudio MLMobile

Eskalate

Product Manager & Backend Web Developer

Visit website

Jan 2023 – Jun 2024Addis Ababa, Ethiopia

Led development of Vertex, an interview-preparation platform, as backend lead across six modules.
Conducted market research with 3,000+ survey responses and 50+ stakeholder interviews to guide product decisions.

Node.jsProductBackend

R&D Group

Odoo Developer Intern

Visit website

Dec 2022 – Mar 2023Addis Ababa, Ethiopia

Collaborated with a team of 12 developers to build ERP modules for enterprise clients.
Improved module stability through debugging and code reviews, reducing bugs by 30%.

OdooERPPython

Hillmark Ethiopia

Full Stack Web Developer

Visit website

Jun 2022 – Aug 2022Addis Ababa, Ethiopia

Completed eight functional requirements within a four-week development cycle using the MEAN stack.
Implemented Test-Driven Development practices and deployed the solution using Nginx.

MEAN StackTDDNginx

05 — Selected work

Projects I've built

A mix of applied-AI research systems and full-stack products. Code links are placeholders — swap them for your repositories.

Healthcare AI · Mobile2025

Tuberculosis Screening Application

End-to-end, AI-powered mobile application that screens for tuberculosis from cough audio recordings — engineered to run offline on low-resource devices.

Built ML pipelines for audio preprocessing, feature extraction, model training, and inference.
Evaluated CNN and Audio Spectrogram Transformer architectures for cough classification.
Optimized lightweight models for offline mobile deployment in low-connectivity environments.

PyTorchAudio MLCNNTransformersMobile

View on GitHub

Full-Stack · Product2023 – 2024

Vertex — Interview Prep Platform

Interview-preparation platform where I served as backend lead across six modules and helped shape product direction from thousands of user signals.

Designed and built backend services across six product modules.
Informed roadmap with 3,000+ survey responses and 50+ stakeholder interviews.

Node.jsExpressPostgreSQLProduct

View on GitHub

Full-Stack · Team LeadJun – Aug 2023

Academic Staff Load Management System

Workload-management platform for academic staff covering workload allocation, research-grant tracking, and finance workflows.

Led a team of six developers across 35 functional requirements.
Developed and tested eight modules end to end.

Full-StackTeam LeadMERN

View on GitHub

Mobile · E-commerceMar 2024

PharmaHub

Mobile e-commerce app connecting pharmacies and customers through delivery and live map-integration services.

Built customer and pharmacy flows with delivery and live map integration.

React NativeE-commerceMaps

View on GitHub

06 — Research

Publications

Peer-reviewed research in multilingual NLP and machine-translation detection.

AcceptedCOLM 2026Conference Paper · 2026

Human vs Machine Translation Detection: A Cross-Model, Cross-Domain, and Low-Resource Analysis

Primary author·Carnegie Mellon University · Languages & Spatial Technologies Lab

A systematic study of how detectable machine-translated (MT) text is from human translation (HT) across high-resource (English, Spanish) and low-resource (Swahili, Afrikaans) languages. We evaluate zero- and few-shot decoder-only LLMs — GPT-5, Gemini 2.5 Pro, and Claude Opus 4 — alongside fine-tuned multilingual encoders (mDeBERTa, XLM-R), and show that input granularity and translation quality are decisive factors for robust MT detection in web-scale corpora.

First systematic evaluation of state-of-the-art decoder-only LLMs (GPT-5, Gemini 2.5 Pro, Claude Opus 4) for MT detection.
Introduces a 120k-example multilingual dataset spanning legislative, conversational, and medical domains.
Shows that chunking inputs to ~300 tokens substantially improves cross-domain and cross-translator generalization.

Multilingual NLPMachine TranslationLLM EvaluationmDeBERTaXLM-RLow-Resource

Read the paper (PDF)View at COLM 2026

07 — Contact

Let's build something

Have a role, a research collaboration, or a project in mind? Send a note and it lands straight in my inbox.

haileyabsera3@gmail.com

Phone

+250 796 162 536

Location

Kigali, Rwanda

GitHub LinkedIn