Skip to content
SASamuel Ajala

Portfolio ’26

Index of work

Samuel Ajala, AI/ML Engineer

AI/ML Engineer building ML and LLM systems that hold up beyond notebooks: production inference APIs, vision models and AI features for existing products, deployed on the cloud.

Status
Full-time & freelance
Focus
LLM apps & integration
Based
Lagos, NG
Selected workBook a callWhatsApp me

(01) About

A model is easy to demo. I make it fast, reliable and cheap to run.

Fig. 1 — Samuel, Lagos

I’m an AI engineer who takes machine learning from the notebook to production, with a background in Electronics and Computer Engineering from Lagos State University.

Most of my work sits where models meet real products and use cases, including inference APIs other teams depend on, LLM features added to existing apps, and vision models that read and process documents. I shape systems output into formats that product teams can use directly.

Then I make it hold up in production. I’ve halved inference latency, kept services at ~99% uptime on cloud and cut infrastructure costs by 55% in a month. Every project ships with documentation, so the next engineer can pick it up without me.

(02) Impact

Faster, cheaper, always on.

Results from ML systems I've shipped to production. I measure where it matters: latency, uptime and cost.

Lower AWS costs within a month, from App Runner and container image optimisation
55%
Pitch deck analysis on Confidential, down from over 2 minutes
<40s
Uptime on AWS App Runner, excluding AWS-wide outages
~99%
Lower inference latency on Daurn AI, from ~60s to ~30s
50%

Case study

Confidential

An ML service that reads startup pitch decks and scores them

  • Upgraded the pipeline to vision-based models with text extraction and probability-based funding rubrics.
  • Cut pitch deck analysis from over 2 minutes to under 40 seconds.
  • Re-architected the service to be inference-only, moving business logic out of the ML layer.
  • Deployed to AWS ECS with automated container deployments from ECR, on right-sized resources.
  • Agreed structured JSON schemas with product owners so scores plug straight into the product.

Across products

Production & data

Keeping inference APIs fast, available and affordable for the teams that depend on them.

  • Maintain production inference endpoints across multiple products.
  • Halved Daurn AI inference latency through model optimisation and system changes.
  • Built scrapers with a storage pipeline and automated CSV export to S3.
  • Cleaned NGX stock market data as the groundwork for time-series forecasting.
  • Every ML project ships with full README documentation, so any engineer can pick it up.

(03) Selected work

Models and products, built and shipped.

Machine learning and LLM projects, from trained models to live products.

  1. 01

    Time-series

    EV Battery RUL Prediction

    LSTM model that predicts the remaining useful life of lithium-ion batteries on the MATR dataset, with incremental fine-tuning so it keeps learning from new cycles.

    • Python
    • TensorFlow
    • LSTM
  2. 02

    LLM product

    Daurn AI

    LLM platform that writes a tailored resume and cover letter for every job application. I halved its inference latency, from ~60s to ~30s.

    • Python
    • LLM
    • GPT
  3. 03

    LLM tool

    AskSpaak

    Ask any webpage a question and get a quick, grounded answer back.

    • Python
    • LLM
  4. 04

    NLP

    Social Media Sentiment Scraper

    Scrapes Nairaland and Reddit comments into a dataset pipeline with automated CSV export to AWS S3, then runs NLP sentiment analysis on them.

    • Python
    • NLP
    • AWS S3
  5. 05

    Computer vision

    CIFAR-10 CNN Classifier

    Convolutional neural network that classifies ten everyday object categories from the CIFAR-10 dataset.

    • TensorFlow
    • Computer Vision

(04) Services

From the first model to production.

End-to-end AI engineering: framing the problem, training the model, and keeping it running in production.

  1. 01

    Custom ML Model Development

    Problem framing, data preparation, training and evaluation of models tailored to your data and business goals.

  2. 02

    LLM & Generative AI Apps

    Chatbots, RAG systems and AI assistants built on modern LLMs, wired into your product through reliable APIs.

  3. 03

    AI Integration for Existing Products

    LLM features added to the product you already have, with structured JSON outputs and inference APIs your team can call.

  4. 04

    Computer Vision Systems

    Vision-based models for classification, detection and document understanding, from prototype to production.

  5. 05

    MLOps & Cloud Deployment

    Containerised deployments on AWS (App Runner, ECS, ECR) or a VPS, with CI/CD, monitoring and cost tuning built in.

  6. 06

    AI Automation & Data Pipelines

    Scrapers, storage pipelines and NLP that turn messy web data into clean datasets in S3.

Toolbox

Modelling
  • Python
  • TensorFlow
  • NLP
  • Computer vision
LLM apps
  • LangChain
  • OpenAI GPT
  • RAG
Serving
  • FastAPI
  • Docker
  • REST APIs
Deployment
  • AWS App Runner
  • ECS & ECR
  • S3
  • VPS
  • CI/CD

(05) Process

Clear steps, Seen at every stage

How a typical engagement runs, whether I'm joining your team or building for you.

  1. 1

    Scope

    We agree on the problem, the data and a measurable definition of done before any model is trained.

  2. 2

    Prototype

    A working baseline early, evaluated on your own data, so you see real results instead of slides.

  3. 3

    Ship

    Production code behind a FastAPI service, containerised with Docker and deployed on AWS or a VPS.

  4. 4

    Optimise

    Monitoring, cost tuning and clear documentation, so the system keeps running well after handover.

Hiring

Full-time roles

I'm open to AI/ML engineering roles where I own AI systems from data and training through to production.

Talk about a role

Building

Freelance & contract

Scoped builds: LLM integration for your existing product, RAG apps or custom models.

Book a call