S H A D H I N L A B

Loading

Generative AI

HR Assistant: generative AI for faster, more consistent screening

How we built a generative AI recruitment assistant that scores and ranks candidates against each role, while recruiters keep the final say.

All case studies
Illustrative product screen for HR Assistant
Illustrative product screen
Client
HR Assistant, our own product
Industry
Recruitment
Services
Generative AI
Platform
Web, works with existing ATS
Status
Live product
Summary

HR Assistant is a generative AI recruitment tool built by Shadhin Lab. It reads job descriptions and resumes, scores and ranks each candidate against the role, and gives recruiters a shortlist to review. It runs on Amazon Bedrock with LangChain, and recruiters make every final decision.

  • Who: HR teams and recruiters hiring at volume
  • Built: AI screening, scoring and ranking
  • Stack: Amazon Bedrock, LangChain, custom scoring
  • Result: faster, more consistent shortlists
01 · The challenge

Too many applications, too little time to review them well

Recruiters face rising applicant volumes, and reviewing resumes by hand is slow and inconsistent. Keyword filters in traditional tools miss qualified people and surface irrelevant profiles.

HR teams also lose time to admin work, which leaves little room for the decisions that matter.

Before
  • Manual resume review for every role
  • Keyword filters miss good candidates
  • Different reviewers, different standards
After
  • Every resume screened as it arrives
  • Matching on skills, experience and context
  • Consistent, auditable scoring
02 · The solution

An assistant that does the first pass, so recruiters can decide

HR Assistant analyses the job description and each resume with large language models, going beyond keyword matching to skills, experience and context.

Custom scoring algorithms rank candidates with transparent scores, LangChain orchestrates the pipeline, and Amazon Bedrock scales it during peak hiring. Recruiters review the ranked list and shortlist.

  1. 1Add the roleUpload or write the job description
  2. 2Collect resumesBy upload, sourcing tools or your ATS
  3. 3AI scoresEach candidate is compared, scored and ranked
  4. 4Recruiter decidesReview the shortlist and invite to interview
03 · Key features

What the platform does

24/7 screening

New applications are evaluated as they arrive, so no candidate waits.

Candidate matching

AI compares skills, experience and context against the role.

Scoring and ranking

Transparent scores highlight the best fits for each position.

Works with your ATS

Fits into existing hiring systems instead of replacing them.

Human in the loop

Recruiters keep final authority over every hiring decision.

Hiring insights

Pipeline and trend data to improve recruitment over time.

04 · Architecture

The technology behind it

Models
LLMs on Amazon Bedrock
Orchestration
LangChainCustom scoring algorithms
Security
AES-256 at restTLS 1.2+ in transitRole-based access
Integration
ATS and HRIS APIs
05 · Results

What changed

Shorter time to shortlist

Screening, matching and ranking happen automatically, cutting manual review.

More consistent evaluation

Every candidate is scored the same way, with results recruiters can audit.

High-volume hiring

Handles large application spikes without adding HR staff.

Privacy-first data handling

Minimal data collection, encryption and role-based access.

Outcomes as described by the product team. No client figures have been published yet.

06 · Questions

Questions about this project

Does HR Assistant replace recruiters?

No. It handles screening and ranking, and recruiters review the results and make every final decision.

Which AI models does it use?

Large language models on Amazon Bedrock, orchestrated with LangChain, plus Shadhin Lab’s own scoring algorithms.

How is candidate data protected?

Data is encrypted at rest with AES-256 and in transit with TLS 1.2 or later, and role-based access limits who can see it.

Get started

Want results like these for your business?

Tell us the problem. We will tell you how we would solve it, and what it would take to run it on AWS.

Not ready to talk? Email hello@shadhinlab.com and an engineer will reply.