I build production software, machine learning systems, and research-driven AI applications.

My work spans reliable AI, computer vision, healthcare AI, ML systems, backend infrastructure, and AI-powered products.

LinkedIn · ORCID · GitHub · Email

I am a Senior Software Engineer focused on building reliable software systems and applied machine learning solutions.

My experience combines:

• Production backend systems, APIs, cloud infrastructure, payment platforms, and distributed software
• Machine learning research in computer vision, healthcare AI, NLP, reinforcement learning, and reliable AI
• AI product development across education, finance, and accessibility
• Research-driven engineering that connects ML experimentation with production systems

Highlights:

• Senior Software Engineer at Municipal Parking Services
• M.S. in Artificial Intelligence from The University of Texas at Austin
• Co-author of a NeurIPS 2023 publication on machine learning reliability
• Author of RibAssist 3D, a medical imaging research system for confidence-aware 3D rib fracture localization
• Founder and builder of MintedBrain, Qubee Kids, and BirrValue
• Former Assistant Lecturer at Addis Ababa University

NeurIPS 2023

Research studying whether language-guided representations can improve the reliability of image classifiers under transient hardware faults.

The approach combines:

• Language-model-generated class descriptions
• CLIP text embeddings
• Vision classification architectures
• Hardware fault injection experiments

Reported results:

• 5.5× average improvement in hardware reliability
• Up to 14× improvement in the most vulnerable layer
• Approximately 0.3% average accuracy reduction
• Compatibility across multiple image-classification architectures

Paper: https://arxiv.org/abs/2311.14062

Code: https://github.com/TalalWasim/TextGuidedResilience

arXiv 2026

RibAssist 3D is a research system exploring confidence-aware localization of rib fractures by combining independently detected findings from orthogonal CT-derived projections.

The work investigates whether AP and lateral detections can be matched across views and selectively reconstructed into reliable 3D fracture locations while maintaining a controlled false-output rate.

Key contributions:

• Biplanar fracture detection from AP and lateral projections
• Rib-side and rib-level anatomical addressing
• Cross-view correspondence using anatomical constraints
• Confidence-gated 3D localization with abstention
• Reproducible case-disjoint evaluation
• Data, model, and evaluation provenance tracking
• Human-in-the-loop review workflow

Key findings:

• Demonstrated accurate geometric localization when reliable cross-view correspondence is established
• Identified lateral-view detection as the primary operational bottleneck
• Established a reproducible framework for confidence-aware biplanar localization

RibAssist 3D is intended as an assistive research system and is not a medical device or standalone diagnostic tool.

Paper: https://arxiv.org/abs/2608.06914

Code: https://github.com/kabJhai/RibAssist-3D

AI learning and discovery platform focused on practical AI education, structured learning paths, tutorials, tools, and real-world workflows.

https://mintedbrain.com/

Interactive Afaan Oromo learning platform designed for children and heritage learners.

Available on Android, iOS, and macOS.

30,000+ downloads

https://qubeekids.com/

Financial information platform for Ethiopian exchange rates, remittance comparisons, currency conversion, and related financial data.

https://birrvalue.com/

Technology initiative focused on building practical software products, intelligent systems, and digital platforms.

https://www.kuushtech.com/

• Building production AI systems and software infrastructure
• Developing reproducible ML evaluation pipelines
• Researching reliable and trustworthy AI systems
• Applying machine learning to healthcare, computer vision, and intelligent systems
• Building AI-powered products for education and accessibility

• Machine Learning Systems
• Reliable and Interpretable AI
• Computer Vision
• Medical Imaging
• Healthcare AI
• Multimodal Learning
• Reinforcement Learning
• Robotics
• AI Infrastructure
• Backend Systems
• Distributed Systems
• AI Product Engineering

Python • C++ • Ruby • Java • JavaScript • TypeScript • SQL • C

PyTorch • TensorFlow • Keras • scikit-learn • OpenCV • FastAI • YOLO • ResNet • Transformers • Computer Vision • NLP

Ruby on Rails • Node.js • Express • React • Angular • Flutter • REST APIs

AWS • Docker • Linux • PostgreSQL • Redis • MLflow • Git • CI/CD

Master of Science in Artificial Intelligence

Completed August 2026

Relevant coursework:

• Machine Learning
• Deep Learning
• Advances in Deep Learning
• Reinforcement Learning
• Natural Language Processing
• Planning, Search, and Reasoning Under Uncertainty
• Online Learning and Optimization
• Optimization
• AI in Healthcare
• Ethics in AI

Bachelor of Science in Software Engineering

December 2020

Very Great Distinction / First Class Graduate

• NeurIPS 2023 Publication
• Fatima Fellowship
• Best Undergraduate Thesis Project Award
• Very Great Distinction / First Class Graduate

LinkedIn: https://www.linkedin.com/in/kabila-haile/

ORCID: https://orcid.org/0009-0008-6740-3214

GitHub: https://github.com/kabJHai

Email: kabilahailesoboka@gmail.com