Shashank Singh logo
HomeResearchProjectsSystemsSkillsExperienceResumeContact
Contact
Shashank SinghAI/ML Developer

AI Engineer building production-ready RAG systems, AI agents, LLM applications, FastAPI backends, and intelligent automation platforms.

RAG SystemsAI AgentsFastAPILLM Apps

Build with me

Production AI, clean interfaces, real systems.

ContactGitHub
Copyright 2026 Shashank Singh. AI engineering portfolio for retrieval, agents, and intelligent automation.
Back to case studies

Research Note 01

ATS Resume AI

An applied RAG system for resume analysis, ATS scoring, and job-fit feedback using local model components and retrieval.

Repositoryats-resume-ai
Live AppVercel
Primary StackFastAPI + RAG
PythonSystem View
ParserATS RulesEmbeddingsFeedback
AI Resume IntelligenceATS Resume AI
Abstract

This project studies a resume-intelligence workflow where retrieval, deterministic checks, and model feedback operate together. The goal is not to replace human editing, but to create a repeatable decision aid for ATS readiness and role alignment.

Research Question

How can an AI resume analyzer provide useful ATS feedback while keeping recommendations grounded in the user's actual resume and target role?

Problem

Resume tools often produce shallow keyword advice. The product question is whether a resume can be parsed, scored, and improved with evidence from role-aware retrieval without inventing candidate experience.

Proposed System

The system combines deterministic ATS checks with retrieval-backed guidance and model-assisted recommendations so the output stays specific, inspectable, and suitable for a job-seeker workflow.

Architecture

Resume Upload
Parser
Rule Checks
Embedding Index
Retrieved Guidance
Model Feedback

Keywords

ATSRAGresume parsingFastAPIgrounded feedback

Methodology

  • Parse resume content into section-level units before scoring.
  • Run ATS-style checks for structure, keywords, readability, and missing signals.
  • Retrieve role guidance using embeddings before generating suggestions.
  • Present recommendations as reviewable edits rather than automatic truth claims.

Evaluation Plan

  • Check whether feedback maps to explicit resume sections.
  • Review suggestion specificity against target job language.
  • Track whether generated bullets preserve factual claims.

Challenges

  • Separating deterministic resume problems from model-generated writing suggestions.
  • Keeping recommendations useful without fabricating achievements.
  • Designing feedback that a user can act on section by section.

Impact

  • Turns resume analysis into a structured optimization workflow.
  • Demonstrates production-minded RAG beyond chatbot interfaces.
  • Shows backend-first AI engineering with practical user outcomes.

Limitations And Future Work

Resume quality is partly subjective and depends on the role.

A model can still overgeneralize if retrieval context is weak.

Add side-by-side diff review.

Add evidence citations for each recommendation.

Introduce role-family benchmark resumes for evaluation.

Explore the implementation

Open the GitHub link or continue through the rest of the AI portfolio.

GitHub LinkAll Projects