Carril Agency
Agency · web + AIA branding, web and growth agency for founders and marketing teams. I work across the web builds and AI automation.
Visit site 
Daniel Iyalekhue
AI / Software Engineer
I build AI that has to work after the demo.
That means engineering the whole path from raw data to a useful decision: retrieval, evaluation, predictive models, APIs and the interface around them.
Open to work · Birmingham, UK danieliyalekhue@gmail.com
Experience
- 2026 — NOW
Software / AI Engineer (Placement) at Blockchain Technology Ltd
Engineering production software and AI features, with an emphasis on maintainable APIs, reliable integrations and shipping usable product increments.
- 2024 — NOW
AI / ML Engineer at Carril Agency
Built an AI reporting-automation pipeline (LangChain + OpenAI over Google/Meta Ads & GA4) on a FastAPI backend, plus an NLP competitor-analysis tool (BERTopic, spaCy) and a RAG brand-content generator.
- 2025
Data Science Intern at Amdari
Built and evaluated ML pipelines (XGBoost, Random Forest, SHAP, LSTM) for churn, HR attrition and predictive maintenance, then turned the results into business recommendations with Streamlit and Power BI.
- 2023 — 24
Web Developer at Carril Agency
Built and maintained client Webflow sites with custom JavaScript, GA4/Meta Pixel integrations and Core Web Vitals improvements; translated Figma designs into responsive experiences.
Awards
- 1st place AI @ BCU Day AI Hackathon Apex Mind took first prize in a six-team challenge with 20 BCU students at STEAMhouse. Context ↗ Certificate ↗19 JUN 2026
- Winner West Midlands @ UKREiiF Next-Gen Challenge Our four-person team won the Birmingham Knowledge Quarter placemaking brief. I brought the AI, engineering and systems-thinking lens to user journeys, accessibility and data-informed decisions. Context ↗ Certificate ↗21 MAY 2026
- 2nd place Oracle APEX Hackathon · BCU Innovation Fest Apex Mind placed second in a low-code, data-driven app challenge focused on Oracle APEX and AI-enabled enterprise solutions. Context ↗ Certificate ↗22 MAY 2026
Stuff I do
AI agents, RAG and document-intelligence systems, ML pipelines and evaluation, full-stack web apps, dashboards, data visualisation, and developer tools.
Websites I’ve shipped
live · 3 builds Visit site Nuclii
Events platformA platform to discover, host and book real-world events, pop-ups and experiences near you.
Visit site hottake
Markets · in progressA markets product, currently being rebuilt for its next version.
Visit site Stack
Testimonials
I've had the chance to see Daniel approach challenges with both creativity and consistency, and it's been impressive to watch.
Daniel is a focused, creative, and dependable person to work with. He has a strong technical mindset and is always looking for better ways to solve problems and improve ideas. What stands out most is his ability to combine creativity with practical execution, especially across technology, web development, AI, and digital projects. I'd highly recommend Daniel to anyone looking for someone hardworking, thoughtful, and easy to collaborate with.
Selected work
all on GitHub →
01 AccessOps COCO AI
Computer Vision · AccessibilityAn image-captioning model that generates alt-text for accessibility. An end-to-end COCO pipeline from CNN+LSTM baselines through transfer learning and RL fine-tuning, with a human-reroute deployment policy and measured BLEU-4 gains at each stage.
DocSage
Document AI · RAGA production PDF RAG system. Upload documents, chunk and embed them locally, then ask questions in a streaming chat with source-grounded citations and persistent vector storage.
DraftDNA
LLM + RAGAn AI writing workspace for academic work. It reads your samples, the brief, the rubric and your approved sources, then produces structured, referenced drafts with citations, confidence labels and DOCX/PDF export.
RepoLens AI
Developer Tools · Local-firstDrop in a codebase ZIP and it scans the stack, runs static analysis and a local-LLM (Ollama) code review, then generates an employer-facing readiness report with deterministic scoring and Markdown export.
AURAFIND
Recommender SystemAn AI fragrance-discovery app. It turns natural-language requests into structured preferences and ranks a 100-fragrance catalogue with transparent scoring, plus filters, a dupe finder and side-by-side comparison.
OPS Platform
Applied AI · B2BThe foundation for an AI manufacturing-operations platform. A B2B dashboard, an RFQ workspace and a quote calculator on a FastAPI and PostgreSQL backend, with SQLAlchemy models and Alembic migrations.
Loan-Approval Bias Audit
Explainable & Responsible AIA loan-approval classifier, audited for fairness across demographic groups using disparate impact and equal-opportunity difference. SHAP explanations, and a responsible-AI writeup of the findings, limits and mitigations.
Lab · live
Small, working demos rather than screenshots. Each one runs in your browser right now.
How an LLM reads text
Language models don't read words, they read tokens. Type below and watch the real GPT tokenizer (cl100k_base) split your text, live, in your browser.
8 tokens44 chars1.14 tok/word
Sampling temperature
How a model picks its next token. Drag the temperature and watch the distribution sharpen or flatten. It's the same control behind “deterministic” and “creative” output.
0.70 · balanced
Contact
Let's build something.
Finishing my MSc in 2026 and open to work: Applied AI / ML Engineer roles, Birmingham, UK. The fastest way to reach me is email.
danieliyalekhue@gmail.com Strengths
- End-to-end ML system design
- LLM engineering & RAG pipelines
- Explainable & trustworthy AI
- Model deployment & containerisation
- Business-focused problem framing
- Communicating AI to non-technical stakeholders