Abdul Basith Mohammed

Data Science Intern · Manulife

Toronto, ON

Built core systems for AskHR — a production RAG chatbot serving employees across 15+ countries — under direct executive oversight, on a 2–3 day release cycle.

• Architected a multi-index hybrid retrieval framework on Azure AI Search, fusing vector and BM25 keyword scoring with semantic reranking, region-hierarchical filtering, and manager-aware chunk pairing.
• Redesigned the preprocessing pipeline by consolidating 6–8 sequential agents into a unified module with async parallel LLM execution — reduced end-to-end query latency from 14s to ~2s (85% reduction).
• Built a hybrid acronym enrichment agent combining hash-based region matching with selective LLM fallback, cutting unnecessary model calls and improving multilingual query grounding.
• Conducted exploratory research and technical scoping for a Predictive Attrition Model, translating stakeholder requirements into a structured PoC roadmap for early flight-risk detection.
• Collaborated in an AI Context Engineering workshop on a natural-language-to-analytics chatbot that translates plain English into structured Pandas workflows.

Associate ML Engineer · Escapade Technologies

Shipped a production ANPR system, built ETL pipelines, and delivered client-facing analytics across multiple verticals.

• Developed and deployed an Automated Number Plate Recognition (ANPR) system, training a license plate object detector that improved plate-read accuracy from ~30% to 97% in production — a 5× increase in correctly detected vehicles.
• Built ETL pipelines (Python, Pandas, SQL, Regex, API integrations) feeding interactive dashboards in Power BI and IBM Cognos for transactional and operations data.
• Optimized data collection and preprocessing workflows using SQL, REST APIs, and Python (Pandas, NumPy, Regex) to improve reliability and processing efficiency.
• Delivered client-facing analytics projects — income prediction, sales performance — applying multiple linear and logistic regression with one-hot encoding, grid search, and k-fold cross-validation, visualized in Power BI and Seaborn.

Data Science Intern · Delta Sigma Technologies

Built an income prediction model on Census Bureau data, paired with interactive analytics dashboards.

• Developed and deployed the 'Average Income Predictor' on Census Bureau datasets, hitting 94% test accuracy.
• Improved model robustness with PCA dimensionality reduction, grid search tuning, XGBoost, and k-fold cross-validation — validated via F1-score.
• Designed and presented interactive dashboards in Matplotlib and Seaborn highlighting patterns and insights from large-scale socioeconomic data.