Makeresume
Build free
MakeResume
M.Sc. Computer Science Research Foundation

Structured Output Synthesis & Evidence-Weighted Scoring

How ongoing M.Sc. Computer Science research at Savitribai Phule Pune University / Maharashtra institutions powers MakeResume's deterministic employability engine.

Deterministic vs Generative Scoring

Unlike conversational chatbots where the same resume gets varying scores upon each prompt, MakeResume uses a mathematical, transparent formula in plain code:

Readiness = [ Σ (wᵢ · eᵢ) / Σ (wᵢ) ] × 100

Must-haves weighted at 3; good-to-haves at 1; evidence evaluated at 1.0 (proven in project) or 0.5 (self-declared).

Schema-Enforced Output Synthesis

For candidates without resumes (Door B), unstructured text entry in Marathi/Hindi is parsed into strict, schema-validated JSON Career Profiles before single-column ATS templates are synthesized.

Guarantees 0% parsing dropout rate on modern ATS parsers (TCS iON, Workday, Greenhouse).

Taxonomy & Synonym Normalization

Over 100 skill synonym mappings account for colloquial variations across tier-2/3 institutions ("MS Excel" vs "Excel", "ReactJS" vs "React", "G-Code" vs "CNC Programming").

Ensures fair evaluation regardless of how a candidate spells their coursework.

Research Focus Areas

  • Information Extraction from Unstructured Resumes: Multi-engine parsing combining pdfjs-dist AST token streams with strict JSON LLM extraction.
  • Evidence Verification Heuristics: Disambiguating active project demonstration from keyword stuffing by inspecting sentence syntax and action verbs.
  • Vernacular-to-Corporate Alignment: Translating Marathi and Hindi vocational terminology into STAR-aligned English professional resumes.
Experience the Research in Action