Aijaz Khan

About

Data science/NLP practitioner with experience at NVIDIA and Microsoft building production-grade NLP and data-linking systems. Has delivered high-performing pipelines (e.g., F1 0.92) and large-scale entity resolution (F1 0.89), plus semantic search using embeddings and Pinecone with ~30–40% relevance gains, backed by rigorous validation (A/B tests, ROUGE, MRR) and strong MLOps/workflow tooling (Airflow, Databricks, FastAPI, MLflow, Prometheus/ELK).

Experience

Education

Key Strengths

• Built and deployed production NLP pipeline for unstructured customer feedback across product lines (BERT/GPT; FastAPI + Airflow/Databricks)
• Strong model evaluation discipline: precision/recall/F1, ROUGE, cross-validation, and A/B testing vs baselines (logistic regression, LSTM)
• Delivered measurable outcomes: F1=0.92 for NLP classification/extraction; reduced manual analysis time via SME-validated topic linking
• Entity resolution at scale across multi-source customer/product datasets using hybrid rules + XGBoost (F1=0.89)
• Implemented semantic search with embeddings + Pinecone; improved relevance ~30–40% and validated with MRR/cosine thresholds
• Production-grade workflow engineering: orchestration (Airflow/Prefect), distributed processing (Spark/Dask), CI/CD (GitHub Actions), monitoring (Prometheus/ELK), testing (PyTest), and ML lifecycle (MLflow)
• Clear prototype-to-production decisioning using technical thresholds (ROUGE-L ~0.88–0.89), latency/stability, and pilot user feedback