Executive Summary
HR Mind is a global resourcing company founded in 2010, offering end-to-end recruitment and HR solutions across eight industry verticals, including infrastructure, renewable energy, IT, healthcare and manufacturing. It serves MNCs, SMEs, startups and joint ventures, and also provides payroll outsourcing, salary analysis and candidate assessment services.
Peritos Solutions developed Recruit Assist, an AI-powered recruitment platform on AWS that automates resume screening, candidate classification, keyword extraction and JD-based ranking using machine learning and NLP. Built on serverless infrastructure, it reduces screening time from days to minutes, minimises bias and fits into HR Mind’s existing applicant tracking workflows.
Project outcomes
Results & Impact
90%
Reduction in manual screening time
< 2 min
Resume classified & ranked vs JD
AI Ranked
Shortlist delivered to recruiter
100s
Resumes processed per job in parallel
Background
HR Mind’s pipeline of job openings was growing quickly, and the number of resumes per role was outpacing what recruiters could process manually. The company needed to automate screening without compromising the quality and accuracy its clients expected.
Recruitment at scale is fundamentally a data-matching problem. Peritos was engaged to build a platform that parses job descriptions, ingests and classifies resumes, matches and ranks candidates, and gives recruiters a central dashboard for the whole pipeline.

Challenge
Each opening attracted hundreds of resumes, most of them irrelevant, and reading every one was the biggest time drain in recruitment. Job descriptions often listed 20 to 40 requirements, making manual matching error-prone.
Different recruiters applied different criteria, and vocabulary differences such as “ML” versus “Machine Learning” caused valid candidates to be missed. With no objective scoring, shortlists were hard to justify to clients.
Delivering a shortlist took two to five days, which cost placements in competitive markets, and past resumes were not being re-matched against new openings.
Solution
Resumes in PDF, Word or text format are stored securely in Amazon S3, which triggers Lambda functions for processing. Amazon Textract extracts text, including from scanned documents, and SageMaker models classify each resume and extract skills, experience, education and job titles.
The matching engine combines exact keyword matching with semantic similarity, weighting hard skills, experience, education, soft skills and location to produce a composite score from 0 to 100. Auth0 handles authentication, and AWS WAF and IAM protect the platform.
Key Solution Capabilities
- JD Parsing: Extract weighted keywords from job descriptions and expand them with synonyms.
- Bulk Resume Processing: Upload 100+ resumes at once for automatic parsing and classification.
- Semantic Matching: Find suitable candidates even when they use different terminology.
- Skills Gap Analysis: See matched, partially matched and missing skills for each candidate.
- Explainable Rankings: Review a transparent score breakdown behind every shortlist position.
- Bias Mitigation: Exclude gender, age and ethnicity from all scoring calculations.
- Client-Ready Exports: Export ranked shortlists as PDF or CSV in one click.
Approach
Peritos delivered the platform in seven phases: discovery, data pipeline, NLP and ML models, matching engine, React.js dashboard, testing and tuning, and go-live. HR Mind recruiters were involved throughout to validate AI output against their own domain expertise.
During testing, AI shortlists were compared with manual shortlists, models were tuned to match HR Mind’s expertise, and a bias audit was carried out. Industry-aware keyword libraries were built for each of the eight verticals, and stored candidates are automatically re-scored against every new job description.
At go-live, Peritos set up CloudWatch monitoring, trained recruiters and provided hypercare support. Lifecycle policies archive older resumes to S3 Glacier to control long-term storage costs.









