AI & Machine Learning AssessmentRESEARCH
Vanco AI Suite
Sales forecasting pipeline, YOLOv8 ASL hand detector, and a Hybrid RAG physics chatbot.
METRIC 01Validation RMSLE of 0.40718 vs Kaggle Public 0.43179 (no time leakage).
METRIC 02ASL Detection precision of 0.938 operating at 30 FPS on CPU.
METRIC 03Grounded RAG tutor refusing out-of-domain queries and returning citations.
Industrial or academic systems require multi-modal pipelines (forecasting, vision, text QA) that must operate reliably under distinct edge constraints.
Attempted Strategy
Initially attempted to split time-series sales forecasting data using random train-test splitting.
Observed Symptom
Severe time leakage caused the validation model to overfit, failing to generalize to public test sets.
Resolution & Fix
Replaced with a chronological validation splitter matching the test set duration, reducing validation RMSLE to 0.40718.
LightGBMYOLOv8FAISSNetworkXFastAPIOpenCVGemini API