Available for opportunities & collaboration

Daniel Data Scientist

AI ENGINEER · DATA SCIENTIST · LLM ENGINEER · MLOPS

I'm an AI Engineer and Data Scientist with 6+ years of experience delivering production-ready AI, machine learning, and data-driven solutions across commercial, healthcare, and enterprise environments. I specialise in designing and deploying LLM, RAG, and Agentic AI systems, by turning raw data into scalable, production-ready intelligence via stakeholder collaboration

STATUS: ACTIVE │ LOCATION: DUBLIN, IRELAND │ ROLE: REMOTE · HYBRID
6+
Years Experience
58+
Projects Implemented
Favourite Tools
PythonLangChainLangGraphPyTorchTensorFlowHuggingFaceScikit-learnPandasNumPyAzureDockervLLMOllamaLlama.cppModalLangSmithLlamaIndexFastAPIStreamlitGitDataBricks
PythonLangChainLangGraphPyTorchTensorFlowHuggingFaceScikit-learnPandasNumPyAzureDockervLLMOllamaLlama.cppModalLangSmithLlamaIndexFastAPIStreamlitGitDataBricks
01 // About

Building intelligence, one model at a time.

Turning raw data into scalable, production-ready intelligence.

I'm an AI Engineer and Data Scientist with 6+ years of experience delivering production-ready AI, machine learning, and data-driven solutions across commercial, healthcare, and enterprise environments. I specialise in designing and deploying LLM, RAG, and Agentic AI systems, by turning raw data into scalable, production-ready intelligence via stakeholder collaboration

My approach blends rigorous experimentation with smooth delivery to bring value to businesses via data-driven insights. Leading to informed decision-making.

print(data-scientist & ai-engineer-profile)
Focus: LLM · Analytics · AI/ML ·RAG · Agentic AI  |  Location: Dublin, Ireland🇮🇪  |  Available for: Remote / Hybrid Roles
02 // Skills & Certifications

Technical Toolkit

Applied AI & ML

NLPSupervised LearningUnsupervised LearningCustomer SegmentationBehavioral AnalyticsPredictive AnalyticsPrescriptive AnalyticsDiagnostic AnalyticsDescriptive Analytics

LLM & Agent Engineering

LLMsLangChainAgentic AIPrompt EngineeringMCPGenAILangGraphLlamaIndexHuggingFace TransformersAxolotlRAG Systems

Programming & Data

PythonPandasNumPyScikit-learnSQLTensorFlowPyTorchPower BIGit

Cloud & MLOps

Microsoft AzureMLOpsDockervLLMOllamaLlama.cppModalRunpodLangSmith

Soft Skills

Stakeholder ManagementMentoringCommunicationAnalytical ThinkingProblem SolvingCross-Functional CollaborationTechnical Documentation

Certifications

Ready Tensor, LLM Deployment & Engineering, 2026Ready Tensor, LLM-Tuning Specialist, 2026Ready Tensor, Agentic AI Engineer, 2025Ready Tensor, Agentic AI Expert, 2025Ready Tensor, RAG Systems Expert, 2025

Languages

English (Fluent)Mandarin (Fluent)
03 // Featured Projects

Selected Projects

portfolio/rag-platform
ERKP● COMPLETED

Enterprise RAG Knowledge Platform

Designed and deployed a production-grade, asynchronous RAG application which reduced legal research time by 40%, to enable lawyers and interns to retrieve relevant legislation information and references from private legal database.

LangChainDeepLake Vector StoreEmbeddingsCohere Re-rankerRAG
portfolio/lorafrica
LFA● COMPLETED

LoRAfrica (LLMs for African History) - LLM Fine-Tuning & Deployment

End-to-end LLM fine-tuning for African history — trained with LoRA and DeepSpeed, quantized with GGUF (llama.cpp, Ollama) and deployed to scalable GPU infrastructure for low-latency inference.

LoRARunpodDeepSpeedGGUFvLLMModalAxolotlLangSmithllama.cppOllamaStreamlit
portfolio/research-qa
RQA● COMPLETED

RAG-Based Research QA Assistant

Retrieval-augmented research assistant answering questions from research papers with cited sources — increasing retrieval speed and accuracy for academic workflows.

LangChainHugging FaceCohereDeepLake
portfolio/agentic-research
ARA● COMPLETED

Multi-Agentic Research Assistant — LangGraph

Agentic AI assistant using LangGraph to orchestrate LLM workflows — auto-generating research abstracts and summarising web content, with tracing, testing and deployment via Streamlit.

LangGraphLLMsStreamlitAgentic AI
04 // Playbooks

AI Delivery Playbooks

Production RAG Deployment
Step-by-step procedure for taking a RAG system from prototype to production — data prep, chunking, retrieval, evaluation and monitoring.
RAGLangChainVector DBvLLM
PB-001 · HIGH
LLM Fine-Tuning with LoRA
End-to-end procedure for instruction fine-tuning a base LLM with LoRA — dataset curation, training, evaluation, quantization and deployment.
LoRAAxolotlDeepSpeedGGUF
PB-002 · HIGH
Agentic AI Orchestration
Procedure for building reliable agents with LangGraph — graph design, tool wiring, tracing, testing and deployment.
LangGraphAgentic AITool CallingLangSmith
PB-003 · MEDIUM
NLP Customer Insight Analysis
Procedure for turning raw customer reviews and interactions into decision-ready insights using embeddings, topic modelling and sentiment analysis.
NLPTopic ModellingSentimentReviews
PB-004 · MEDIUM
Segmentation & Forecasting
Procedure for delivering segmentation, risk classification and demand forecasting solutions that survive contact with real business processes.
ClusteringForecastingClassificationSQL
PB-005 · MEDIUM
05 // Case Studies

Client Case Files

Enterprise RAG Deployment — Legal CLOSED
Production-grade, asynchronous RAG platform for a legal firm — 40% faster legal research via DeepLake retrieval, Cohere reranking and citation-backed answers served over REST APIs.
RAGLangChainDeepLakeCohereGemini
CS-001 · HIGH · 2025-06
Retail Fashion Churn Analytics — End-to-End CLOSED
Led end-to-end churn analytics for a retail fashion client across descriptive, diagnostic, predictive and prescriptive stages — pinpointing delivery delays as the main churn driver and improving retention by 14% after validating faster delivery via A/B testing.
ChurnA/B TestingRetailAnalyticsPrescriptive
CS-002 · MEDIUM · 2025-01
Medallion ETL Pipeline — Enterprise Data Engineering CLOSED
Developed and optimised ETL pipeline workflows to ingest, clean and transform 200k+ records using a Medallion (Bronze→Gold) approach — cutting manual effort by 35% and improving data accuracy.
ETLMedallionData EngineeringData QualityAutomation
CS-003 · HIGH · 2023-03
Fraud Detection & Prevention — Banking CLOSED
Analysed behavioural and transaction data for 500K+ high-risk, high-value customers using classification and segmentation models to identify fraud — cutting fraudulent payments and complaints by 14% and financial losses by 20%.
Fraud DetectionSegmentationClassificationRiskCustomer Analytics
CS-004 · HIGH · 2025-03
LLM Fine-Tuning for Financial Document Summarisation CLOSED
Led LLM fine-tuning and rigorous benchmarking for financial document summarisation — achieving ROUGE-1/ROUGE-2/ROUGE-L scores of 0.51/0.34/0.47 and 78% numeric accuracy, which increased stakeholder confidence.
LLM Fine-TuningBenchmarkingFinancial NLPROUGEEvaluation
CS-005 · HIGH · 2023-05
Healthcare KPI Dashboards — Power BI CLOSED
Developed insightful Power BI dashboards for a healthcare client, tracking operational KPIs, performance trends and service delivery metrics — enabling stakeholders to drill into root causes and improve outcomes by 15%.
Power BIDashboardsHealthcareKPIsRoot Cause
CS-006 · MEDIUM · 2024-06
06 // Publications & Insights

Published Insights

Design of Gas Cyclone Using Hybrid Particle Swarm Optimization Algorithm
Peer-reviewed research applying a hybrid PSO algorithm to optimise gas cyclone design - published in MDPI.
Research · 2021-10 · 15 min
Honorary Award of Excellence — Best International Student Anti-Epidemic Pioneer
Recognised by Liaoning University of Engineering and Technology for outstanding contribution as an international student during the pandemic response.
Recognition · 2020-09 · 5 min
Lessons from Shipping RAG into Production
What actually breaks when RAG leaves the notebook - retrieval quality, evaluation, observability and stakeholder trust.
Insights · 2026-02 · 8 min
LoRA Fine-Tuning: What Actually Moves the Needle
Practical notes from fine-tuning LLMs with LoRA - data quality, target modules, and why evaluation beats iteration.
Insights · 2026-01 · 10 min
Why CRISP-DM Still Matters for MLOps
A 25-year-old methodology is still the backbone of reproducible, governable AI delivery.
Insights · 2026-07 · 7 min
07 // Contact

Let's Connect

Interested in working together on AI, data science, or MLOps? I'm always open to new opportunities and conversations.