Abubakar Aliyu

AI & Data Engineer | Applied AI, Cloud & Data Systems

AWS Certified Solutions Architect – Associate

Paris, France

explore my portfolio

Who I Am

I'm an AI & Data Engineer with a background in Computer Engineering and Data Science, focused on building intelligent, reliable, and cloud-native systems that solve practical problems.

My experience spans AI engineering, data engineering, cloud platforms, and MLOps, from developing production data pipelines and cloud applications to building retrieval-based AI systems, fine-tuning language models, and evaluating their reliability. I'm particularly interested in grounded AI, reasoning, and the challenges of making intelligent systems dependable in real-world settings.

I bring an analytical mindset, adaptability, and a strong sense of ownership to my work. I enjoy investigating complex problems, learning new technologies, and finding practical solutions. My approach combines critical thinking with sound engineering practices, including modular design, automation, testing, and continuous improvement.

I believe effective engineering is not always about choosing the most complex solution, but about understanding the problem, making thoughtful trade-offs, and building something that works reliably and is maintainable over time. I also value collaboration and learning from different perspectives.

Outside of engineering, I enjoy puzzle games, sports, and participating in hackathons.

By the Numbers

Click to explore each section

Years in AI & Data

Core Skills

Featured Projects

Degrees

Certifications

Awards

Experience

Building solutions that make a difference

Featured Projects

Applied AI, Cloud & Data Systems

Applied AI & Intelligent Systems

Retrieval-Augmented Reasoning with Self-Correction for Biomedical QA (2026)

▼

Type: Master's Thesis

Goal: Investigate Retrieval-Augmented Reasoning Modeling (RARE), using retrieved evidence and structured reasoning-focused training, and extend the framework with error detection and self-correction; evaluated on biomedical question answering.

Role & Team: Team Lead (4 members) - implemented the RARE baseline, coordinated research, and integrated error detection and self-correction.

Tools & Methods: Llama-3.1-8B, QLoRA, structured knowledge/reasoning tokens, reasoning-weighted loss, FAISS, SciBERT, Flan-T5, PyTorch, Redis, Hugging Face Inference Endpoints, LLM evaluation.

Outcome & Impact: Implemented the RARE baseline and integrated a self-correction workflow; achieved 0.826 error-detector ROC-AUC and 0.82 correction BERTScore F1. Evaluated reasoning reliability and deployed GPU inference with caching.

0.826 detector ROC-AUC 0.82 correction BERTScore F1 4-person research team

GitHub Repository Dataset on HF Full Thesis Details →

Data-Centric LLM Fine-Tuning with Systematic Evaluation for Domain Name Generation (2025)

▼

Type: Applied AI Project

Goal: Fine-tune Llama-3.1-8B-Instruct for creative, brandable domain generation and improve model quality through targeted data curation and systematic evaluation

Role & Team: Individual project - diagnosed training-data quality issues, curated targeted examples, designed a systematic quality assessment framework, and compared model iterations through controlled experiments

Tools & Methods: Llama-3.1-8B-Instruct, QLoRA, PyTorch, Hugging Face Transformers, Weights & Biases, LLM-as-a-Judge, systematic quality scoring, Hugging Face Spaces

Outcome & Impact: Identified training-data quality issues and LLM-as-a-Judge bias. Reduced generic training patterns from 38.0% to 27.8%, improved extended safety performance from 64% to 82% (11 test cases), and deployed the fine-tuned model on Hugging Face Spaces.

99.1% success rate 82% extended safety 27.8% generic patterns

GitHub Repository Live Demo Model on HF Dataset on HF

Production ML Systems

Telecom Customer Churn — ML Data Pipeline & Microservices Platform (2025)

▼

Type: Academic Team Project

Goal: Develop an end-to-end ML platform to predict customer churn for a telecom use case, addressing a reported 43% quarterly churn rate through automated data pipelines, model serving, and monitoring.

Role & Team: Team Lead (5 members) - architected microservices system, implemented FastAPI model serving, designed automated data pipelines with Airflow

Tools & Methods: Docker, FastAPI, Streamlit, PostgreSQL, Apache Airflow, Grafana, scikit-learn, microservices architecture

Outcome & Impact: Built a containerized ML platform integrating Airflow orchestration, PostgreSQL, FastAPI prediction serving, and Grafana monitoring, with automated data-quality checks and batch prediction workflows.

GitHub Repository

ChatJourno (2024-2025)

▼

Type: AI Product Project originating from the CJID AI-in-Journalism Fellowship

Problem & Goal: In West African newsrooms, journalists face a 15-minute window to update breaking news with credible backstories. Under this pressure, research often comes from unreliable sources. The project aimed to empower journalists with an AI-assisted fact-checking system.

Role & Team: AI Engineer – co-developed the initial MVP during the CJID AI-in-Journalism Fellowship (May-Nov 2024) and continued working with the team afterward to develop ChatJourno toward a standalone product. Developed the Python/LangChain backend, integrating data collection, retrieval, and LLM-assisted fact-verification workflows for the MVP.

Tools & Methods: Python, LangChain, BeautifulSoup, Streamlit, web scraping, retrieval workflows, dataset curation from trusted sources

Outcome & Impact: Co-developed a functional AI-assisted fact-checking MVP, integrating retrieval and LLM-based verification workflows. Presented ChatJourno at the 2024 Media Innovation West Africa Conference and contributed to subsequent product testing and refinement.

Featured in TheCable ↗

Cloud & Data Engineering

Serverless AWS Data Pipeline - NYC Taxi Data (2025)

▼

Type: Academic/Personal Project

Goal: Build a fully event-driven, scalable pipeline for processing and analyzing large volumes of transportation data

Role & Team: Individual contributor - architected event-driven ETL pipeline, implemented PySpark transformations, configured fine-grained IAM permissions

Tools & Methods: AWS Lambda, AWS EMR, AWS Glue, AWS Athena, AWS CloudFormation, PySpark

1.4M+ records processed Serverless architecture

Outcome & Impact: Built an event-driven AWS data pipeline processing 1.4M+ NYC taxi records, using PySpark for transformations, automated orchestration, and IAM-based access controls.

GitHub Repository

Scalable AWS Analytics System - Student Performance Analytics (2025)

▼

Type: Academic/Personal Project

Goal: Build a fault-tolerant, scalable platform for analyzing student performance data with interactive visualizations

Role & Team: Individual contributor - designed AWS architecture with custom VPC, implemented Streamlit dashboard, configured Auto Scaling Groups

Tools & Methods: AWS (VPC, EC2, ALB, Auto Scaling, S3, IAM), Streamlit, Python data science stack

2-4 auto-scaling EC2 instances 70% CPU threshold scaling

Outcome & Impact: Built an AWS-based analytics pipeline to ingest, transform, and analyze student performance data, producing structured datasets and visualizations to support educational insights.

GitHub Repository

Technical Skills

Core expertise and technical capabilities

AI & Machine Learning

PyTorchHugging FaceScikit-learnTransformersLLM Fine-TuningLoRA/QLoRARAGLangChainLangGraphFAISSNLPComputer VisionLLM EvaluationWeights & BiasesFastAPI

Cloud & DevOps

AWS (Lambda, ECS, S3, EventBridge)AWS (EMR, Glue, Athena, IAM, VPC)AWS CloudFormationGCP Cloud RunDockerGitHub ActionsCI/CDLinuxGit

Data Engineering & Analytics

PythonSQLSnowflakePostgreSQLPySparkApache AirflowETL/ELTData QualityData WarehousingGrafanaStreamlit

Education

Academic foundation in AI and engineering

Master of Science in Computer Science (Data Science and Analytics)

EPITA School of Engineering and Computer Science – Paris, France

Completed: Sept 2026

Research Focus: Retrieval-Augmented Reasoning, self-correction, and systematic LLM evaluation

Thesis: Retrieval-Augmented Reasoning with Self-Correction for Biomedical QA. Led a 4-person team to implement RARE, which externalizes retrieved knowledge so model training can focus on reasoning rather than knowledge memorization, and extended the framework with systematic error detection and self-correction. Evaluated the approach experimentally using biomedical question answering as the application context, with BERTScore, LLM-as-Judge, hallucination detection, and systematic ablation studies. Deployed the resulting system on a dedicated Nvidia A10G GPU with a custom handler, Redis caching, QLoRA quantization, and scale-to-zero.

GitHub Repository Dataset on HF

Bachelor of Engineering in Computer Engineering

Ahmadu Bello University – Zaria, Nigeria

Completed: Feb 2022

Grade: Second Class Honours (Upper Division), CGPA: 3.60/5.00

Research Focus: Network Systems and Routing Protocols in Intermittent Connectivity Environments

Thesis: Enhanced PRoPHET Routing Protocol with Buffer Management Techniques in Delay-Tolerant Networks

Methods: Implemented and evaluated PRoPHET routing with four buffer-management algorithms (MOFO, DLA, DL and FIFO) through controlled network simulation experiments using ONE Simulator.

Outcome: Compared buffer-management strategies under simulated Delay-Tolerant Network conditions to evaluate their effect on routing performance.

Certifications

Continuous learning and professional development

AWS Certified Solutions Architect - Associate (SAA-C03)

2026

Completed

Certified in AWS architecture design, compute, storage, networking, security, and cost optimization best practices.

Verify on AWS

Google Certified Data Analyst

2023

Completed

Comprehensive certification covering data collection, transformation, analysis, visualization, and data-driven decision making.

Verify on Coursera

Google Certified Project Manager

2023

Completed

Professional certification in project management fundamentals, including planning, execution, and stakeholder management.

Verify on Coursera

Awards & Recognition

Scholarships, fellowships & competitive selections

EY Open Science AI & Data Challenge 2026 | Top 5, France Leaderboard

EY Open Science AI & Data Challenge

Apr 2026

Achieved a top-five position on the French leaderboard among 350+ participants in a machine learning challenge focused on predicting river water quality in South Africa.

Petroleum Technology Development Fund (PTDF) Overseas Scholarship Award

EPITA: École d'Ingénieurs en Informatique

Sep 2024

Awarded a competitive postgraduate scholarship by the Petroleum Technology Development Fund (PTDF) to pursue postgraduate study abroad, based on academic merit and selection through the national scholarship process.

AI-in-Journalism Fellow

Centre for Journalism Innovation and Development

May 2024 - Nov 2024

Selected from 500+ applications across Nigeria and Ghana for CJID's AI-in-Journalism Fellowship. The fellowship supported the development of ChatJourno, which was presented at the 2024 Media Innovation West Africa Conference.

Summer Institute in Computational Social Science (SICSS-Calabar)

Duke University, Social Science Research Council, Mathematica

September 18 - 29, 2023

Selected to participate in SICSS-Calabar, part of the international Summer Institute in Computational Social Science program, for intensive training in computational methods for social science research.

TotalEnergies/NNPC National Merit Scholarship Award

Ahmadu Bello University

Mar 2018

Awarded through the TotalEnergies/NNPC National Merit Scholarship scheme in recognition of academic excellence during undergraduate study.

Languages

International communication capabilities

English

Bilingual Proficiency

French

Intermediate (B1)

Currently seeking: AI/ML Engineer, Data Engineer, Research Engineer, Software/Cloud Engineering, and Data & AI Consulting roles

We're a Match

If you're building reliable AI, cloud, or data systems, I'd love to hear about it. Feel free to reach out.

Email WhatsApp Instagram