Software & applied AI engineerLondon, UK / Portfolio 2026

Hello, I’m Olusegun Ekoh.

Engineeringintelligence.

I turn complex ideas into dependable software. From AI agents and RAG pipelines to the systems that keep them running in the real world.

Explore my work Let’s talk
The engineer behind the systemsOlusegun Ekoh
Fig. 01 / Olusegun EkohHuman, first.
6+Years building
2M+Users served
MScAI & Robotics · Distinction
A few things I’ve built

01 / Selected projects

Ideas, put to work.

Practical experiments in AI, retrieval, and systems built to hold up in production.

Project 01 / Applied AI · Infrastructure

Doc Insight

A multi-tenant Retrieval-Augmented Generation (RAG) service for document Q&A. Upload text files or PDFs, including scanned ones, and ask questions answered from their content, with parsing and embedding handled asynchronously by a separate worker.

PythonFastAPIPostgreSQLpgvectorRedisOpenAI
Explore the code
Under the hood
  • Async ingestion: the FastAPI API validates and returns 202, while a separate worker parses and embeds documents
  • Postgres-backed job queue (SELECT ... FOR UPDATE SKIP LOCKED) with lease-based crash recovery and typed retries
  • Text extraction with pypdf, falling back to OCR (PyMuPDF + Tesseract) for scanned pages
  • Semantic search over 1536-dimensional OpenAI embeddings with PostgreSQL + pgvector
  • Multi-tenant API key auth, with every document and query scoped to its owner
  • Per-key rate limits, plus Redis-cached query embeddings and answers
  • Prometheus metrics and alerts, OpenTelemetry traces from upload to worker, and logs in Loki

Project 02 / Machine learning · Information retrieval

Semantic Search

A semantic search engine that retrieves research papers based on meaning rather than keywords, processing 41K ML papers using transformer embeddings and GPU-accelerated similarity search.

PythonPyTorchSentence-TransformersFaissCUDA
Explore the code
Under the hood
  • High-quality 768-dimensional embeddings using MPNet model
  • GPU-accelerated indexing via Faiss for near real-time search
  • Embedding generation in ~41 seconds on GPU (~30 iterations/second)

02 / Behind the work

A little context.

Good systems start with good questions.

More about my background

I’m a software engineer working at the intersection of applied AI and dependable infrastructure. I care as much about what happens after deployment as I do about the first working prototype.

Over 6+ years, I’ve built banking applications serving more than two million users, payment platforms, and AI systems for aerospace documentation. Today, my focus is on agents, retrieval, and turning language models into useful, measurable products.

My foundation is in Physics & Electronics, with a First Class degree and an MSc in Artificial Intelligence & Robotics, awarded with Distinction. That mix shapes how I work: curious about the problem, rigorous about the solution.

What I bring

Systems thinking. Production discipline.
A practical approach to AI.

03 / Capabilities

The working toolkit.

The right tools for the problem, from the first API to the production pipeline.

01

AI/ML

Machine LearningDeep LearningNLPPyTorchscikit-learnAI AutomationRAG PipelinesAgentic AI
02

Frontend

TypeScriptReactNext.jsTailwind
03

Backend & APIs

PythonFastAPITypeScriptNode.jsExpressNestJSJavaSpring BootMicroservices
04

Cloud & Infrastructure

AWSPostgreSQLMongoDBRedisDockerKubernetesTerraform
05

DevOps & MLOps

CI/CDOpenTelemetryPrometheusGrafanaModel DeploymentCost Optimization

04 / The journey so far

Built through experience.

Mar 2026 — Present

Agentic Labs HQ

London, UK · Current

Applied AI Engineer

Developed a multi-tenant RAG platform for document search and question answering, providing source-grounded responses with tenant-isolated access to uploaded knowledge.

Contributions & outcomes
  • Engineered reliable ingestion for text and scanned PDFs using OCR, asynchronous workers, crash recovery and idempotent writes, recovering from failures without duplicating document data.
  • Implemented distributed tracing, correlated logs and CI-tested alerting across the API and workers; verified incident diagnosis through failure drills and document-level traces.
  • Established a 100-question benchmark for retrieval and answer quality, compared LLM judge assessments against 50 blindly hand-graded answers and added a CI gate to catch retrieval regressions.

May 2025 — Feb 2026

University of Hertfordshire / Sono Vision Ltd

UK

Applied AI Engineer

Designed an AI-assisted technical writing tool to help aerospace and defence writers produce ASD-STE100 Simplified Technical English while preserving technical meaning.

Contributions & outcomes
  • Built a modular Python/FastAPI service combining spaCy-based rule validation, deterministic correction and LLM-assisted rewriting for sentence-level ASD-STE100 compliance.
  • Implemented safeguards for AI-generated revisions through protected-data checks, rule revalidation and local semantic analysis, with human review for uncertain or safety-related content.
  • Implemented PostgreSQL persistence with transactional audit records, version-aware translation caching and optimistic locking to preserve traceability and prevent conflicting reviewer updates.
  • Established automated evaluation against 944 reference sentence pairs and Prometheus/Grafana monitoring to track translation quality, response latency and LLM costs.

Jan 2022 — May 2024

Oval Finance

Delaware, USA

Product Software Engineer – Backend & Platform

Built distributed backend services and internal administrative products for blockchain and traditional payments, spanning Kafka event streaming, provider integrations, KYC systems and operational workflows.

Contributions & outcomes
  • Delivered end-to-end internal admin dashboards, building APIs and user interfaces for Operations and Compliance teams to manage payments, customer accounts and KYC workflows.
  • Developed a secure, idempotent KYC synchronisation microservice on a Kafka pipeline, storing documents in Amazon S3 and PostgreSQL, with scheduled reconciliation and time-limited presigned URLs for Compliance review.
  • Designed a unified payment-provider integration service using hexagonal architecture, centralised secrets management and provider-specific retry and error-handling policies.
  • Supported services in production on AWS, resolving incidents using CloudWatch and working with Operations and Compliance teams to restore service and prevent recurrence.
  • Led architecture and code reviews, mentored engineers and established shared standards for API design, error handling, testing and service design.

Oct 2020 — Feb 2022

Guaranty Trust Bank

Lagos, Nigeria

Full-Stack Software Engineer

Built customer-facing banking applications, backend APIs and automated operational workflows for mobile and web platforms serving more than two million users.

Contributions & outcomes
  • Improved dashboard load performance through server-side pagination, leaner DTOs, Redis caching and versioned APIs, contributing to a 35% increase in adoption.
  • Designed responsive user interfaces and integrated them with backend APIs to deliver customer-facing account and transaction workflows.
  • Automated bulk salary-upload and interbank transfer workflows using modular validation, asynchronous processing and exception routing, reducing manual processing time by 45%.
  • Eliminated peak-time transaction history timeouts by replacing N+1 queries with a joined query and adding a composite database index, reducing endpoint latency by 40%.

Mar 2019 — Aug 2020

3Cloud Limited

UK

Software Developer

Built and maintained RESTful API endpoints, delivering features from ticket to production.

Contributions & outcomes
  • Wrote unit and integration tests with JUnit and Mockito for new and existing services, raising code coverage for owned modules.
  • Investigated and resolved production bugs through log analysis and query debugging.
  • Documented API endpoints and deployment procedures, reducing onboarding time for new team members.

05 / Education & credentials

A foundation in curiosity.

2024—2026Distinction

MSc Artificial Intelligence & Robotics

University of Hertfordshire, UK

Thesis: Automated Translation from Standard English to Simplified Technical English (ASD-STE100) using Prompt-Engineered LLMs.

2012—2016First Class

BSc Physics and Electronics

Redeemer's University, Nigeria

Strong foundation in Electronics, Computational Physics, and Mathematical Modelling.

Professional certifications

Google Cloud Certified Professional Cloud Developer

06 / Writing & thinking

Notes from the process.

All writing

30 min read

Building a Semantic Search Engine with Transformers and FAISS

Learn how to build an efficient semantic search engine using sentence-transformers and FAISS to search through 41,000+ ML research papers by meaning, not just keywords.


4 min read

Hello World - Welcome to My Engineering Blog

Welcome! I write about AI/ML engineering, full-stack development, and building intelligent systems at scale.

07 / Start a conversation

Have a good
problem to solve?

A new project, an interesting challenge, or an opportunity to build something useful. I’d love to hear about it.

olusegun.oe@gmail.com
GitHub LinkedIn