Fatima Firdouse

I’m an AI and Data Science enthusiast focused on solving real-world problems with data—building end-to-end systems that combine data pipelines, machine learning, and user-focused applications.

I’ve worked with large-scale datasets (500k+ records), designing and automating ETL workflows on AWS using Python, Pandas, and NumPy. I also engineered domain-specific features and trained a Random Forest churn model, then translated outputs into concise executive summaries for senior stakeholders.

Across my projects, I’ve prioritized impact over complexity: a dual-flow ETL/ML pipeline that scores candidate-job matches and runs bias detection, plus a production retrieval pipeline with multi-stage ranking (BM25 + semantic fused via RRF, followed by cross-encoder reranking). I’m comfortable working in ambiguous environments by breaking problems down into measurable steps and iterating toward clear outcomes.

Production quality matters to me, so I integrate observability and evaluation directly into the workflow. With Langfuse tracing and Ragas evaluation, I’ve captured retrieval and generation latency in live dashboards and achieved strong automated performance (including a Context Precision score of 1.0), while delivering fast, publicly accessible APIs on AWS.