Devanshi Sharma
Backend Engineer
Remote (global)
LinkedIn profile available to registered employers
Self-assessed English levels
AI-Powered Job Search Automation Platform
Cut daily job-screening from ~40 minutes to under 2 minutes (~6.5 hours saved over 2 weeks) while increasing job discovery volume 40x. Architected a polyglot microservice platform: Spring Boot (REST APIs, Thymeleaf UI, Flyway-managed Postgres schema); Python FastAPI as the AI worker running Ollama (llama3.1) for structured resume parsing from PDF/DOCX via Apache Tika., persisted to PostgreSQL. Automated multi-portal discovery via JobSpy across LinkedIn and Indeed, ingesting 526 live postings through a 9-term × 7-location matrix, deduplicated to 451 unique roles on normalized company + title and served through a drill-down explorer. Engineered a 4-component matching engine over pgvector — cosine similarity on persisted 384-dim JD embeddings (44%), tier-weighted skill overlap (28%), YOE-decay experience fit (17%), BFSI domain fit (11%) — scoring 500 JDs in ~5 seconds via batched inference plus single-SQL similarity, ranked best-first per location bucket with visible per-job reasons.