Amel J.

Life‑Cycle Bayesian Forecasting Tool (inprogress, inproduction): – Designing and shipping a continuous Bayesian forecasting system for short life‑cycle products, where standard time‑series methods fall short due to limited history and rapid demand shifts. – Built the full stack end‑to‑end: Databricksjobs for ingestion, preprocessing, and Bayesian model training and inference; a Scala backend for orchestration and serving; and a user‑facing tool deployed on Azure App Service. – Forecasts refresh continuously as new sales data arrives, with posterior uncertainty surfaced to planners alongside point estimates. Mitigation System (anomaly detection + causal explanation + recommendation): – Built an end‑to‑end ML system to detect anomalous shrink, identify likely causes, forecast future losses, and recommend high‑ROI mitigation actions. – Enriched the model with external context by scraping demographic and crime data, and integrated environmental and operational factors. – Delivered an interactive dashboard with "what‑if" scenarios so customers could simulate the expected impact of prevention actions.

• • Designed and shipped Kestea, an AI product that automates LinkedIn content, outreach, and engagement for sales professionals. Live with early customers.

• • Built LLM workflows tuned to each user's voice, replacing weekly manual posting and per‑recipient message personalization.

• • Architected the full serverless AWS stack: Lambda, Amplify, DynamoDB, Cognito, SES, SNS, CloudWatch.

• • Designed cloud infrastructure for managing ATM operational data on AWS for a banking client, including migration from legacy on‑prem servers.

• • Architected multi‑account environments with AWS Control Tower and Organizations, and automated deployment of VPC, IAM, RDS (PostgreSQL), S3, and security groups via CloudFormation and Bash.

• • Developed Glue‑based ETL pipelines integrating with REST APIs, with CodeCommit for CI.