Recruitment operations & data pipelines

Vantage

High-stakes institutional workflows benefit from deterministic, reviewable data transformation rather than opaque extraction alone.

Status: In progress
Domain: Data Systems · Backend Engineering · Experimentation
Stack: Python · Flask · PostgreSQL
Role: Data pipeline · Data model · REST APIs · Experimental evaluation

Open repository

Overview

Vantage consolidates academic-record ingestion, eligibility verification, drive management, and placement tracking into an auditable platform for engineering institutions.

Problem

Placement teams often coordinate irregular spreadsheets, duplicated records, and manual eligibility checks. Small parsing errors can propagate into decisions that affect students and recruiters.

Data / Inputs

CSV and Excel T-Sheet exports are supported by the implemented parser.
Incoming columns are mapped to a canonical schema, with conflicts staged for review.
Duplicate content is detected using a file hash before ingestion.

Architecture / Approach

Flask application with PostgreSQL, SQLAlchemy, and Alembic.
Three-stage parse, normalize, and ingest pipeline.
Conflict states preserve questionable records for explicit admin resolution.
Eligibility rules return both a decision and human-readable rejection reasons.

My contribution

Student-record processing pipeline and normalized PostgreSQL model.
REST APIs and role-aware operational workflows.
Comparison of deterministic parsing with LLM-based extraction on clean and noisy inputs.

Limitations

PDF parsing remains planned rather than implemented.
The migration set and automated test coverage require further production-readiness work.