Fransi Ayele
AI/ML Engineer·Addis Ababa, Ethiopia

Building intelligentsystems thatship to production.

AI/ML Engineer specializing in production machine learning, Generative AI, reliable ML infrastructure and intelligent multi-agent systems.

Currently

AI/ML Engineer @ INSA

Focus

Generative AI · MLOps · Multi-agent systems

  • 0%

    Model iteration speedup

    Kubeflow & Vertex AI on K8s

  • 0%

    Incident reduction

    Monitoring & drift detection

  • 0.0%

    Production uptime

    TensorFlow Serving (p99 <150ms)

  • 0%

    Fewer hallucinations

    Multi-agent LangGraph baseline

  • 0+ GB

    Daily telemetry streaming

    Apache Beam & Kafka

PythonSQLBashC++PyTorchTensorFlowScikit-learnXGBoostSHAPFeature EngineeringModel EvaluationDrift DetectionLLMsRAGLangChainLangGraphHugging Face TransformersLoRAPEFTPineconeDockerKubernetesKubeflowMLflowVertex AIAmazon SageMakerPythonSQLBashC++PyTorchTensorFlowScikit-learnXGBoostSHAPFeature EngineeringModel EvaluationDrift DetectionLLMsRAGLangChainLangGraphHugging Face TransformersLoRAPEFTPineconeDockerKubernetesKubeflowMLflowVertex AIAmazon SageMaker
Portrait of Fransi Ayele Mengesha

Based in

Addis Ababa, Ethiopia

9.03° N · 38.74° E

01About

AI/MLEngineerwith2+yearsofexperiencebuildingproductionmachinelearningsystemsforsecuritytelemetry.SpecializeinMLOps,Kubernetes-basedmodelservingandreliableMLdeployment.Cutmodeliterationtime35%andcontributedtoa25%reductioninproductionincidents.DevelopRAGandmulti-agentLLMapplicationswithLangGraph,LangChainandPinecone.PursuinganMScinArtificialIntelligenceatAddisAbabaUniversity,expected2028.

Education

  • MSc · Artificial Intelligence

    Addis Ababa University

    Sep 2025 – Expected 2028 · In Progress

  • BSc · Computer Engineering

    Addis Ababa Science and Technology University

    Sep 2019 – 2024

Certifications

  • IBM RAG and Agentic AI Professional Certificate

    IBM

  • Generative AI with Large Language Models

    DeepLearning.AI / AWS

  • Machine Learning Specialization

    DeepLearning.AI / Stanford Online

  • Oracle Cloud Infrastructure 2023 AI Certified Foundations Associate

    Oracle

02Capabilities

A full stack forproduction machine learning.

From data pipelines and training to Kubernetes serving, monitoring and agentic LLM applications. Every tool here has shipped in a real system.

  • 04

    Programming

    • Python
    • SQL
    • Bash
    • C++
  • 08

    Machine Learning

    • PyTorch
    • TensorFlow
    • Scikit-learn
    • XGBoost
    • SHAP
    • Feature Engineering
    • Model Evaluation
    • Drift Detection
  • 08

    Generative AI

    • LLMs
    • RAG
    • LangChain
    • LangGraph
    • Hugging Face Transformers
    • LoRA
    • PEFT
    • Pinecone
  • 11

    MLOps & Cloud

    • Docker
    • Kubernetes
    • Kubeflow
    • MLflow
    • Vertex AI
    • Amazon SageMaker
    • Amazon S3
    • Amazon EC2
    • Amazon ECR
    • Cloud Build
    • CI/CD
  • 08

    Data Engineering

    • Apache Beam
    • Apache Kafka
    • Apache Airflow
    • Dataflow
    • Pandas
    • NumPy
    • PostgreSQL
    • BigQuery
  • 04

    Computer Vision

    • YOLOv8
    • ResNet
    • U-Net
    • OpenCV
03Experience

Where the workhas actually run.

Production engineering across high-throughput security telemetry, distributed model serving on Kubernetes, and teaching the next cohort of ML engineers.

  1. 2024 – Present

    AI/ML Engineer

    Information Network Security Administration (INSA)

    Addis Ababa, Ethiopia

    Production ML for streaming security data

    • 01

      Cut model iteration cycles 35% by building Kubeflow and Vertex AI pipelines on Docker and Kubernetes.

    • 02

      Sustained 99.5% production uptime and p99 inference latency under 150 ms using TensorFlow Serving on Kubernetes.

    • 03

      Cut deployment lead time from 2 days to 4 hours by automating ML builds and releases with Docker, Cloud Build and validation gates.

    • 04

      Processed 100+ GB of security telemetry daily using Apache Beam, Dataflow, Kafka, SQL and Pandas.

    • 05

      Reached 89% precision and 81% recall on streaming security data with anomaly detection and classification models.

    • 06

      Contributed to a 25% reduction in production incidents through monitoring, drift detection, model versioning and SHAP explainability.

    • 07

      Aligned engineering, research and analyst teams on ML solutions through documented architecture decisions.

    DockerKubernetesKubeflowVertex AITensorFlow ServingApache BeamDataflowKafkaSQLPandasSHAP
    Next role
  2. 2023 – Present

    Machine Learning Instructor (Part-Time)

    Bright Learn Academy

    Addis Ababa, Ethiopia

    • 01

      Teach supervised and unsupervised learning to 30+ students through practical Python examples.

    • 02

      Train students in data preparation, exploratory analysis and feature engineering using NumPy, Pandas and Scikit-learn.

    • 03

      Guide practical projects in model training, evaluation and optimization.

    PythonNumPyPandasScikit-learnModel Optimization
04Selected work

Systems withmeasured outcomes.

Multi-agent research, enterprise retrieval and medical imaging. Each project is reported with the benchmark it was evaluated against.

FeaturedSeptember 2026—

A production-grade multi-agent autonomous research framework delivering grounded, multi-perspective intelligence briefs with guaranteed schema integrity.

  • LangGraph
  • Gemini
  • FastAPI
  • Pydantic
  • Tavily
  • Docker
Retrieval-Augmented Generation2023—

Enterprise-scale knowledge retrieval system indexing unstructured documents for sub-second, highly factual query resolution.

  • Python
  • LangChain
  • Pinecone
  • GPT-4
Computer Vision2023

Clinical-grade computer vision pipeline segmenting gliomas on multi-modal MRI scans with uncertainty estimation.

  • PyTorch
  • ResNet50
  • U-Net
  • Medical Imaging
Computer Vision2023

Agricultural diagnostic vision model identifying pathologies across 54 distinct crop disease classes.

  • TensorFlow
  • PyTorch
  • CNNs
  • OpenCV
05What I do

Four waysI can help.

Advisory and hands-on engineering for teams taking machine learning from notebook to production.

  • Kubeflow, Docker and Kubernetes pipelines automating model build, release, and low-latency inference serving.

    • Automated CI/CD with validation gates and drift alerts
    • TensorFlow Serving & Triton inference engines (<150ms p99)
    • SHAP explainability, model registry and versioning
06Contact

Let's buildsomethingintelligent.

Open to engineering roles, architecture advisory and collaborations in production ML, agentic AI and MLOps.

francyayele@gmail.com
GitHub LinkedIn CV Addis Ababa, Ethiopia · UTC+3

No newsletters, no spam. Just a reply.