AI ML Engineer with less than a year in LLM Fine-tuning & RAG Pipelines
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Junior AI/ML Engineer with a foundation in statistics, specializing in building and deploying intelligent systems grounded in real-world data experience. Trained at the National Bureau of Statistics, Nigeria, on national-scale datasets, developing rigorous data quality and statistical thinking skills that directly inform model development.
Ladoke Akintola University of Technology (LAUTECH)
Bachelor of Technology (B.Tech) · Statistics
September 1, 2019 – September 1, 2024
Jay Academy
Volunteer Data Science Instructor
June 1, 2026 – Present
India
Codveda
ML Intern
January 1, 2026 – March 1, 2026
India
National Bureau of Statistics Nigeria
Internship Trainee
July 1, 2023 – February 1, 2024
Ibadan North, Oyo State, Nigeria
TinyLlama LoRA Fine-Tuning – LLM Instruction Tuning & Deployment
June 10, 2026 – Present
Enabled domain-specific instruction-following capability in TinyLlama by designing, curating, and formatting a custom 1,000-sample AI Engineering Q&A dataset, then applying LoRA adapters via the PEFT library to fine-tune the model with minimal compute overhead, achieving a strong reduction in training loss across 500 steps. Delivered a fully functional parameter-efficient fine-tuning pipeline by configuring LoRA adapter layers on a base model with no prior task-specific knowledge, producing a fine-tuned model capable of accurate, domain-grounded responses to AI Engineering prompts. Made the fine-tuned model publicly accessible for real-time inference without requiring any local GPU setup by deploying the LoRA-adapted model to Hugging Face Spaces with a Gradio interface for interactive Q&A.
View ProjectWhatsApp AI Health Assistant — RAG + FastAPI + Twilio + Groq
June 10, 2026 – Present
Eliminated hallucinated responses in a live health Q&A system by building a full RAG pipeline that retrieves relevant context from PDF health documents before passing enriched prompts to a Groq-hosted LLM, ensuring document-grounded answers for real WhatsApp users. Delivered WhatsApp as a fully operational conversational AI interface by integrating Twilio's WhatsApp API with a FastAPI backend to receive, process, and respond to user health queries in real time, tested end-to-end. Kept the system lightweight and production-ready by routing inference through Groq's high-speed LLM API instead of self-hosting a model, reducing both response latency and infrastructure cost without sacrificing answer quality.
View ProjectStudent Segmentation Web App - KMeans Clustering + Streamlit
June 10, 2026 – Present
Identified three distinct student behavioral profiles academically focused, moderately engaged, and at-risk by applying KMeans clustering to student lifestyle, study habit, and academic performance data, enabling educators to design targeted intervention strategies. Made unsupervised ML insights accessible to non-technical educators without requiring any code knowledge by deploying the segmentation model as a live, interactive Streamlit web application where users can directly explore cluster profiles. Ensured clustering accuracy across features of varying scales by applying StandardScaler normalization before fitting KMeans and validating the optimal cluster count using the Elbow Method, preventing high-range variables from dominating distance calculations.
View ProjectTrashNet Waste Classification CNN Computer Vision
June 10, 2026 – Present
Restored model accuracy from a near-random baseline to a strong-performing level by diagnosing a generator shuffle misalignment that was corrupting label-prediction pairing in the training pipeline, resolving the data bug before making any architecture changes. Built a waste image classifier across six categories validated by per-class precision, recall, and F1-score by designing a three-layer CNN (Conv2D, BatchNorm, MaxPooling, Dropout) trained on the TrashNet dataset using TensorFlow and Keras. Reduced unnecessary training time significantly by implementing EarlyStopping, ReduceLROnPlateau, and ModelCheckpoint callbacks, halting training early while preserving the best-performing model weights and controlling overfitting.
View ProjectCultural Fit Analysis
The candidate's involvement in volunteer teaching and open-source deployment (Hugging Face Spaces) indicates a collaborative and community-oriented mindset. The diversity of personal projects, from LLM fine-tuning to computer vision and traditional ML, shows a broad interest in AI/ML applications. However, the lack of team-based project experience outside of internships might suggest a need for development in large-scale collaborative engineering environments. The target role of AI ML Engineer aligns well with the candidate's project focus and stated interests.
Soft Skills & Operational Fit
The candidate demonstrates strong initiative and a proactive approach through personal projects and volunteer work. Their experience as a data science instructor suggests good communication and mentorship skills. The ability to work on diverse projects (health assistant, waste classification, student segmentation) indicates adaptability and problem-solving capabilities. However, the limited professional experience (internships) means operational fit in a fast-paced, senior engineering environment needs further validation.