John Lenehan

Senior AI/ML Engineer | Data Scientist

5+ years building analytical and ML systems across SaaS and manufacturing.
Taking models from POC to reliable production systems.

Tech Stack

PythonSQLBash

MCP (FastMCP)RAGHugging FaceOllamaVector DBs

TensorFlow / KerasScikit-learnOpenCVXGBoostStatsmodels

GCP (Vertex AI, Cloud Run)Azure ML StudioCI/CDDockerMLflow

Developing enterprise GenAI solutions using robust RAG/MCP pipelines and domain-specific models.

Engineering scalable ML pipelines and automated CI/CD deployments on cloud platforms.

Building and optimising high-performance production ML models, managing the complete lifecycle from POC to deployment.

Rigorous statistical analysis and experimental design to optimise business-critical processes.

Softworks|Apr 2025 – Present|Dublin, Ireland

Overhauled legacy chatbot system by engineering a production RAG pipeline for vector search across document embeddings, dropping latency by 98% (5+ mins to <3s) and achieving 97% response accuracy.
Architected and deployed a Dockerised MCP server to interface with an enterprise chatbot, achieving <300ms latency while applying access controls to block unauthorised API calls.
Developed and deployed multivariate time series forecasting systems to predict staff availability and demand, identifying coverage dips and achieving MAPE of <3% against observed values.
Established multi-stage CI/CD pipelines to automate deployments across Dev/Test/Prod environments, integrating continuous testing and PR-controlled promotion.

Intel Corporation|Aug 2020 – Oct 2024|Leixlip, Ireland

Resolved demand-shock domain shift in factory time series forecasting by developing discrete event simulation models to synthesise new datasets, achieving MAPE of <5% against observed values.
Achieved >96% recall on foundry defect classification by transferring and fine-tuning an established 14nm CNN to assist labelling, bootstrapping training of a foundry-native CNN model.
Produced statistical models to determine the optimal strategy for a highly sensitive EOY factory shutdown, limiting material loss to <0.15% and ensuring Q1 output targets were met.
Reduced critical product abort rates by 66% on key toolset through rigorous diagnostic analysis of tool telemetry data, root-causing the issue to misaligned tool configurations.
Led a cross-functional team through new product certification for Intel Foundry's first customer products, delivering the final milestone 1 week ahead of commit dates.

Medtronic|Jan 2018 – Aug 2018|Galway, Ireland

Implemented modular stent assembly stations, reducing maintenance downtime from 1 week to 2 hours.
Designed new cleanroom tools to prevent stent damage, reducing stent defect rate by 75%.