Ahmed Malik

• Fine-tuned a BERT-based NER + text-classification pipeline in Python to flag accident-related events across approximately 2 hours/day of pre-transcribed police-radio logs, reaching 92% precision and 89% recall • Deployed a serverless ETL stack (AWS S3, Glue, Lambda) that cut the load time for weekly road crash files from 45 minutes to under 3 minutes, allowing analysts to flag high-risk corridors during the same shift

• Redesigned a five-year digital marketing database (~300 K rows) to Third Normal Form, removing 92% of duplicate rows and reducing median PostgreSQL query latency from 4.7 seconds to 1.9 seconds • Ran K-Medoids in Python to identify high-value segments, lowering annual ad spend by $700 K/year and lifting conversion 1.5× • Created four interactive Tableau dashboards (conversion funnel, cohort health, campaign ROI, and executive KPI scorecard) with 30-minute auto-refresh via Athena/Glue, adopted by VP

• Refactored PostgreSQL stored procedures that load 30 million USDA SNAP program records, shrinking the window from 3.2 hours to 2 hours and stabilizing CRUD performance in an Oracle APEX front end

• Collaborated with faculty to enhance the R programming curriculum by co-authoring six auto-graded hands-on labs, strengthening students’ practical skills in statistical programming • Mentored more than 35 graduate students in regression analysis and statistical inference

• Formulated Logistic Regression propensity models combined with Recency–Frequency–Monetary (RFM) based segmentation to pinpoint high-value customer cohorts for multichannel marketing campaigns amounting to $54 K and obtained a 30% ROI • Conceptualized and analyzed A/B tests on website pages using R, SQL, and Google Analytics, cutting bounce rate by 68% and informing UX improvements • Forecasted monthly lead volume with SARIMA and Facebook Prophet, improving budget-planning accuracy by 25%