Natnael Daba
Verified Expert in Engineering
AI Engineer and Developer
Tucson, AZ, United States
Toptal member since November 4, 2025
Natnael is an AI researcher and engineer with expertise in computer vision, deep learning, and large language models (LLMs). He has built and deployed machine learning systems for image understanding, LLM evaluation, and geospatial analysis. Natnael's current work explores vision-language models that bridge visual and textual intelligence for real-world applications.
Portfolio
Experience
- Python - 6 years
- Machine Learning - 6 years
- Amazon Web Services (AWS) - 6 years
- AI Research - 5 years
- Natural Language Processing (NLP) - 5 years
- Statistical Methods - 5 years
- Deep Learning - 4 years
- Geolocation Detection - 3 years
Preferred Environment
Linux, Cursor AI, Visual Studio Code (VS Code), PyTorch, Jupyter Notebook, Git, Docker, Weights & Biases, NVIDIA CUDA, Slack
The most amazing...
...thing I’ve developed is a cross-view video geolocation system that matches ground-view driving videos with satellite imagery using deep learning and 3D vision.
Work Experience
Graduate Research Assistant
Integrated Sensing and Processing Lab
- Built an end-to-end pipeline for GPS-denied localization from ground video, integrating image retrieval, 3D reconstruction (DUSt3R/COLMAP), BEV rendering, and trajectory estimation for map alignment.
- Designed a prior-guided localization framework that leverages road topology, landmarks, and adaptive sequence windows to enhance retrieval and pose estimation. Improved cross-route localization stability under varying frame rates.
- Developed diffusion-based colorization models to bridge infrared and visible modalities, enabling cross-spectral retrieval and enhancing multi-sensor geo-localization accuracy in low-light or obscured conditions.
Graduate Research Assistant
UCF Center for Research in Computer Vision (CRCV)
- Developed CNN-based classifiers for small-object recognition in satellite imagery, designing reproducible data pipelines and robust quantitative evaluation protocols to ensure consistent model benchmarking.
- Proposed and implemented a novel classifier adaptation framework for changing class priors during deployment, improving test-time calibration and classification accuracy by up to 15% under distribution shift.
- Built a 3D reconstruction and LiDAR point cloud matching algorithm for drone-based scene localization, improving sub-10-meter positioning accuracy by 13% under noisy measurement conditions.
Machine Learning Engineer
mDoc
- Designed and deployed a Dialogflow-based chatbot that reduced clinician response times by over 20%, improving patient communication efficiency.
- Developed and fine-tuned knowledge bases for FAQ systems using advanced data augmentation and transfer learning to enhance answer accuracy and coverage.
- Integrated the FAQ chatbot into web and mobile platforms, enabling over 2,000 patients to access real-time automated support and significantly improving service availability.
Experience
Analysis of Benchmark Contamination in Open-source LLMs
https://github.com/nate-daba/benchmark-contaminationI designed reproducible pipelines for evaluation, contamination detection, and statistical testing, implementing techniques such as guided prompting and TS-Guessing to measure memorization versus genuine reasoning. Experiments used PyTorch, transformers, bash automation, and multi-GPU evaluation scripts for large models up to 70 billion parameters.
The companion toolkit automates contamination detection through bootstrap significance testing and structured logging, enabling researchers and organizations to quantify data leakage risks and enhance benchmark integrity and model transparency.
Education
Doctorate Degree in Electrical and Computer Engineering
University of Arizona - Tucson, AZ, USA
Master's Degree in Electrical and Computer Engineering
Carnegie Mellon University - Pittsburgh, PA, USA
Bachelor's Degree in Electrical and Computer Engineering
Addis Ababa University - Addis Ababa, Ethiopia
Skills
Libraries/APIs
TensorFlow, PyTorch, LLMSanitize
Tools
Jupyter, ChatGPT, Git, Slack
Languages
Python, C, C#, Bash
Paradigms
Automation, Distributed Computing
Platforms
Ubuntu, Amazon Web Services (AWS), Azure, Android, Linux, Visual Studio Code (VS Code), Jupyter Notebook, Docker, Weights & Biases, NVIDIA CUDA
Other
Machine Learning, Natural Language Processing (NLP), AI Research, Deep Learning, Statistical Methods, Geolocation Detection, Code Review, Interviewing, Task Analysis, Source Code Review, Artificial Intelligence (AI), AI Agents, Agentic AI, LangChain, AI Automation, Generative Artificial Intelligence (GenAI), Hugging Face, Image Generation, Stable Diffusion, Multimodal GenAI, Retrieval-augmented Generation (RAG), Vector Databases, Technical Hiring, Digital Signal Processing, Digital Communication, Wireless Communication, Computer Vision, Research, Large Language Models (LLMs), Deep Reinforcement Learning, Transformers, Prompt Engineering, Multi-GPU Training, Data Analysis, Statistical Analysis, Model Evaluation, Reproducible Pipelines, Cursor AI
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring