Transforming HRTech
Transforming HRTech Intelligence for Recruitment Smart Technologies (UK)
Client
Recruitment Smart Technologies (RSTL) is a leading UK-based HRTech company building advanced AI-driven recruitment platforms. Their products require deep data analytics, intelligent automation, and enterprise-grade AI models to understand candidate behaviours, match talent to roles, and deliver measurable hiring efficiency.
RSTL partnered with Brown Earth Technologies UK to enhance three key capabilities:
- AISS – Artificial Intelligence Services Stack for deep behavioural analytics
- Data Science & Model Development for scalable insights
- LLaMA-3 LLM Testing & Issue Categorisation to improve accuracy of model responses
1.AISS – Artificial Intelligence Services Stack for deep behavioural analytics
A fully generated micro-database of households and individuals, representing:
•Age, gender, income, education
•Household composition
•Lifestyle markers and preferences
•Micro-location attributes (postcode-level)
Technology: SDV, probabilistic modelling + UK Census / regional statistics.
2.Data Science & Model Development for scalable insights
Brown Earth delivered a full Data Science Development Framework including:
•Designing the data models
•Infrastructure setup
•Training RSTL team on usage
•Model testing & calibration
This formed the backbone for scalable analytics across RSTL’s ecosystem.
3.LLaMA-3 LLM Testing & Issue Categorisation to improve accuracy of model responses
Objective
RSTL engaged Brown Earth to evaluate and improve the accuracy of LLaMA-3 based responses for their talent-intelligence use-cases.
Work Completed
1) Testing of 20 Client-Flagged Cases
Brown Earth tested all 20 cases RSTL provided and categorised every error type.
2)Partial-Match Accuracy Assessment
Accuracy was calculated with partial matches considered valid (as previously agreed), giving a realistic representation of the model’s capability.
3)Issue Classification
Each test case was mapped to one of the following issue categories:
•Data Ambiguity
•Contextual Misinterpretation
•Missing Knowledge
•Classification Drift
•Structural Output Misalignment
4)No Fine-Tuning Applied
The dataset remained unchanged at RSTL’s request.
Therefore, model accuracy reflected the best achievable result without fine-tuning.
5) Testing Report Delivered
A full testing report was provided to RSTL summarising:
•Case-by-case evaluation
•Accuracy score under partial-match logic
•Categories of failure
•Recommendations for future optimisation
•Potential fine-tuning pathways
Executive Summary
Brown Earth Technologies UK played a pivotal role in strengthening Recruitment Smart’s HRTech platform by delivering:
✔ AISS Deep Analytics Engine – Advanced sentiment, behavioural, and historic data intelligence.
✔ Data Science Development -A fully structured analytics ecosystem ready for scale.
✔ LLaMA-3 LLM Testing & Issue Categorisation -A clear, benchmarked view of model accuracy and actionable recommendations.
Together, these initiatives have helped RSTL create a more intelligent, highly accurate, and scalable AI recruitment platform—positioning them strongly in the competitive HRTech landscape.
