Purpose: We are looking for an expert to benchmark, identify edge cases, and refactor our existing vehicle codifier engine. The primary goal is to increase accuracy to 95% for Top-3 candidates — meaning the correct value should appear within the top 3 predictions with at least 0.95 probability.
Key Responsibilities: Benchmark current codifier engine performance and identify weak points/edge cases. Implement and refine fuzzy matching, embeddings, and LLM-based reranking. Refactor engine components to improve scalability, maintainability, and accuracy. Develop APIs and evaluation endpoints using FastAPI. Provide structured performance reports and optimization suggestions.
Requirements: Strong expertise in fuzzy string matching, semantic search/embeddings, and LLM reranking. Proven experience with FastAPI for backend service design. Knowledge of data-driven evaluation (precision, recall, F1, Top-K accuracy). Familiarity with insurance/vehicle classification domain is a strong plus but not mandatory.
Deliverables: Refactored codifier engine with documented benchmarks. Edge case test suite for regression validation. Evaluation report showing ≥95% Top-3 accuracy on validation set.
