Agentic workflows

Agentic LLM Evaluation Workflow for Resume Screening

A LangGraph workflow evaluates resume–job fit through typed extraction, gap analysis, deterministic score calibration and recommendation generation, with LangSmith tracing.

PythonLangGraphPydanticLangSmithLLM APIs / multi-provider setup

Ask about this project View repository

Choose your depth

Architecture, choices & results

The problem

Reduce ambiguity in resume–job fit analysis by separating structured LLM reasoning from application-owned calibration.

System architecture

Follow the flow from input to evaluated output. Select a stage to inspect its role.

PDF / DOCX inputs — PDF / DOCX inputs

Engineering decisions

Engineering decision

Separate extraction from calibrated scoring

Problem: Raw LLM judgments and unconstrained responses can make scoring logic ambiguous.
Approach: Use typed Pydantic extraction, explicit gap-analysis/post-processing/recommendation stages, deterministic calibration and LangSmith tracing.
Trade-off / scope: Qualitative consistency is supported; measured quality, hiring outcomes and production deployment are not established.

Evaluation, in context.

No quantitative evaluation results are documented.

Reliability & deployment

A documented safety and scope boundary.

Typed outputs and deterministic calibration support qualitative consistency claims; no benchmark dataset size, accuracy, precision, recall, F1 or agreement score is supplied. No validated hiring accuracy, bias reduction, recruiter productivity gain, ATS equivalence or production hiring outcome is established. The baseline workflow is completed; deployment remains in progress and must not be presented as a production service.

Deployment status

Deployment is in progress; no production service claim is supplied.

Context limited to Resume Screening

Challenge this project.

Ask about the decisions, evaluation or boundaries behind the work.

Deterministic answers from documented portfolio evidence · project-scoped

A claim is only the beginning

Where did I use it?
What did I build?
How did I measure it?

Follow the projects, decisions and evaluation behind each answer. Every important claim has somewhere to go.

Direct · measured implementation Supporting · documented use Related · adjacent experience

An inspectable claimPG / 001

CAPABILITYRetrieval-augmented generation

PROJECTAutoCare Agent

IMPLEMENTATION

BM25+semantic retrieval

Trace the evaluation behind the number