Mikiyas Zenebe Bekele

Self-taught AI Engineer and Python Devloper | Machine Learning

+251995641212 mikilezen@gmail.com

scroll down

ABOUT

Hi, I’m Miki, an IT student and a self-taught AI enthusiast who loves building things with code. I’m currently studying Information Technology at Arsi University, and in my free time, I teach myself Machine Learning, Deep Learning, and Computer Vision through online courses, documentation, and hands-on projects. I enjoy experimenting with AI from face recognition and emotion detection to building chatbots, recommendation systems, and language tools. Most of what I know comes from curiosity, consistency, and solving real problems through small personal projects. I believe learning by doing is the best way to grow, and I’m always working on something new: improving my skills, exploring new frameworks, and building projects that help people or make tech more accessible. My goal is to become a strong AI engineer who creates meaningful and practical solutions for real-world challenges.

EXPERIENCE

AI Engineer & Machine Learning Developer Designed, trained, and deployed machine learning and deep learning models using scikit-learn and PyTorch for prediction, classification, and recommendation systems. Built AI-driven recommendation engines and implemented large-scale data processing workflows using PySpark to improve performance and optimization. Developed intelligent LLM-powered applications with LangChain, integrating custom pipelines, retrieval modules, and context-based reasoning for automation and chatbot development. Implemented computer vision systems using OpenCV and PyTorch, including face recognition, emotion detection, and various image-analysis solutions. Worked on NLP, audio processing, and translation projects, leveraging Hugging Face Transformers to build and fine-tune models for text understanding, generation, and multilingual applications. Contributed to multiple AI projects involving model design, dataset preparation, evaluation, and deployment.

PROJECTS

AILVision - AI-Powered CCTV & Face Recognition System
AILVision is an AI-powered security system I am developing to modernize the existing CCTV infrastructure at Arsi University. The system integrates real-time face recognition, intelligent monitoring, and automated alerts, transforming traditional cameras into a smart AI surveillance platform.
Key Features
Real-Time Face Recognition Uses deep learning–based facial embeddings and FAISS vector search to identify people quickly and accurately. Automated CCTV Analysis Converts raw CCTV footage into meaningful insights using computer vision models. Attendance & Access Tracking Can be extended to track student/employee attendance, restricted-area access, and visitor logs.
Live Monitoring Dashboard Displays recognized individuals, unknown visitors, detection logs, and system analytics.
Scalable Architecture Designed to work across multiple CCTV cameras with a Python + Flask backend and optional cloud deployment.
Tech Stack
Python, OpenCV, face_recognition, InsightFace Flask, FAISS, NumPy, Pandas
Berta Language Machine Translation
See on Wikipedia
This project focuses on building a low-resource machine translation system that translates between Berta (Bertha) and English. The goal is to contribute to the digitization and preservation of Ethiopian indigenous languages using modern Natural Language Processing (NLP) techniques.

Data Collection & Preparation

Collected, cleaned, and aligned 6,000+ parallel Berta–English sentence pairs. Performed normalization, tokenization, and preprocessing to ensure high-quality training data. Addressed challenges of low-resource languages such as limited corpora, irregular spellings, and morphological variation.

Model Architecture

Built a Seq2Seq (Sequence-to-Sequence) neural architecture with: Custom Encoder–Decoder structure Attention mechanism for improved contextual alignment Experimented with multiple architectures and finalized a Transformer-based model optimized for low-resource scenarios. Achieved meaningful translation results despite the low-resource environment, demonstrating feasibility for underrepresented languages. Trained the model using PyTorch and Hugging Face Transformers.

Deployment

The model is published and open-source on Hugging Face:
Model Repository: https://huggingface.co/Mikile/Bertha-translation-encoder
This allows researchers, linguists, and developers to use or fine-tune the model for further work in Berta NLP. Goal & Impact This project’s mission is to: Preserve and promote Berta language through modern machine learning Provide a starting point for future Ethiopian indigenous language NLP research Support tools like translation systems, chatbots, digital dictionaries, and educational apps Reduce the technological gap for underrepresented languages
AI Recommendation System
Implemented both content-based and collaborative filtering methods. Designed systems to deliver personalized recommendations using machine learning techniques.
LangChain-Based Chat Assistant
Built an intelligent chatbot powered by LangChain and LLMs. Integrated custom knowledge-retrieval pipelines for more accurate and context-aware responses.

SKILLS

CERTIFICATIONS

I’m always excited to connect with fellow AI enthusiasts, collaborate on innovative projects, or discuss the latest trends in technology. Whether you have a question, an idea, or just want to say hello, feel free to reach out!