Deep Purple AI Consulting

Deep Purple AI delivered an AI consulting and discovery engagement for an Irish precision engineering firm based in Munster with approximately 45 employees. The company designs, manufactures, and services custom mechanical components for pharma, medtech, and food processing clients. Revenue was tied directly to headcount. Engineers were spending significant time on repetitive documentation, compliance checking, and specification review instead of billable design work.

Deep Purple ran a structured Enterprise Ireland Digital Discovery: on-site visits, stakeholder interviews, workflow mapping, data landscape assessment, and a comprehensive AI opportunity assessment. The engagement identified 5 specific AI use cases ranked by estimated ROI.

Use cases covered: AI-assisted document processing and specification review (largest opportunity, projected 6-8 hours saved per engineer per week), automated compliance checking against building regulations, project reporting automation, submittal and RFI pre-screening, and internal knowledge search across completed project documentation.

Deliverables:

• AI readiness assessment,
• 5 prioritised use cases with ROI estimates and feasibility ratings,
• implementation roadmap (quick wins through to long-term transformation),
• budget estimates per use case with grant funding pathways,
• data readiness assessment,
• executive presentation to leadership.

For the highest-priority use case: We tested sample documents against NLP models to validate feasibility before projecting ROI. Not generic benchmarks. Observed workflow samples and real data.

The discovery stands alone. No obligation to proceed.

Full case study: deeppurple.ai/case-studies/ai-discovery-engineering

Deep Purple AI delivered a predictive quality analytics project for a production facility within one of Europe's largest food groups, employing approximately 120 people at the Connacht site. The engagement was initiated by the group's Head of R&D. BRC and Bord Bia Origin Green certified. 350+ product lines. The company collected 150+ production attributes per batch (raw material specifications, processing parameters, environmental conditions, time-temperature profiles) but had no way to link them to finished product quality outcomes.

Barry Gough (CTO, MSc Machine Learning UCD) led the project with Deep Purple's data science team. Systematic approach: data preparation and feature engineering from time-temperature profiles, R-squared correlation matrices, machine learning models (XGBoost, random forests, support vector machines, explainable boosting machines), SHAP explainability analysis, and quartile analysis comparing top vs bottom performers.

Deliverables:

• clean reusable data schema
• R-squared correlation matrices
• feature importance rankings identifying the attributes that actually drive quality
• quartile analysis
• trained predictive ML models with cross-validation
• SHAP explainability on every prediction
• working prototype prediction tool for the quality team
• technical report and executive summary

Key finding: a small subset of attributes drives the majority of quality variation. The client now knows which measurements matter and which they could reduce or eliminate. Predictive models achieved strong accuracy for primary quality dimensions.

All data processed within EU/EEA. No production data used to train public models. Data returned or securely deleted at completion. Entire engagement required less than 8 hours of client team time across 4 weeks.

Phase 2 scoped for production-grade tooling.

Full case study: deeppurple.ai/case-studies/predictive-quality-food-manufacturing

Deep Purple AI built a custom AI quoting system for an Irish building services contractor with 60 employees and 25 field technicians servicing commercial HVAC, mechanical, and electrical systems across Leinster. The company was producing 30-50 quotes per week. Complex quotes (chiller replacements, planned HVAC overhauls, multi-trade maintenance contracts) took 1-2 hours each. Reactive callouts took 20-30 minutes. Estimating knowledge was concentrated in two people, one approaching retirement.

We built an AI-powered quoting engine that analyses 5+ years of historical ERP job data, finds similar completed jobs, and generates itemised draft quotes with parts, labour, margin, and a confidence score. Human-in-the-loop: every quote is reviewed and approved by the estimating team before sending. The system pre-populates method statements and safety plans from templates.

The AI connects to the existing ERP via read-only access. Works with Sage, SAP Business One, and Microsoft Dynamics. All data processing on EU/EEA cloud infrastructure. GDPR compliant. Full audit trail on every quote. Client data never used to train public AI models.

Results (measured over 6 weeks post-deployment):

• complex quotes from 1-2 hours to 15-20 minutes.
• Reactive quotes from 20-30 minutes to under 10.
• Quote turnaround from 24-48 hours to same day.
• 15-20 hours per week of admin time freed across the estimating team.
• Pricing consistency improved from 15-20% variance to data-driven baseline.

Started with €1,250 AI Discovery (Enterprise Ireland, 80% funded). Build phase 50% grant funded through EI Digital Process Innovation.

Production system. Used daily. Retiring estimator's knowledge now captured in the system.

Full case study: deeppurple.ai/case-studies/ai-automated-quoting-field-services

Deep Purple AI built a custom AI and computer vision platform for an Irish construction contractor with 30-40 field operatives across sites in Ireland, the UK, and Europe. The company needed to replace paper dockets and manual measurement reconciliation. Verified remeasurements showed 15-20% variance. No real-time visibility across sites.

Phase 1: custom mobile app (React Native, Android and iOS) using OpenCV and ArUco marker calibration to calculate area from photographs. Operatives photograph completed work, draw an outline, and submit. Two minutes. ±3% accuracy. Works fully offline. GPS and timestamps on every submission.

Phase 2: full operations platform with real-time dashboards across all active construction sites. Per-operative productivity and earnings tracking. Weather integration correlated with output. Anomaly and overlap detection. Multi-language support. Natural language query interface where managers ask questions in plain English and get data-backed answers.

Phase 3: machine learning quality detection trained on labelled data from Phase 2. Structured client-facing reports for government and infrastructure contracts.

Tech: React Native, Node.js, Python, OpenCV, PostgreSQL, Google Cloud (EU, GDPR compliant). Custom software owned by the client.

Results:

• paper dockets eliminated.
• ±3% accuracy (was 15-20%).
• Real-time visibility across all sites.
• GPS-stamped, photo-evidenced, auditable records.
• Client billing from verified data.
• Back office reviews quality same day.
• Progress reports build themselves with every submission.

Project was 50% funded through Enterprise Ireland. Full case study with grant pathway details on our website.

Production system. Used every day on real Irish and European construction sites.

Full case study: deeppurple.ai/case-studies/computer-vision-construction