Amel J.

I help retail and e-commerce teams forecast demand, detect anomalies, and build the supporting data and cloud infrastructure end-to-end. Three years full-time at a retail consultancy across the full demand spectrum (long life-cycle staples, seasonal categories, and short life-cycle fast-fashion or new-launch items), a delivered Upwork engagement with a fast-fashion e-commerce client, a 9-month AWS infrastructure mission for a banking client, and a serverless AWS product I co-founded and built end-to-end. What I solve: - Demand forecasting for established retail catalogs. Time-series at scale with trend and seasonality decomposition, contextual drivers (promotions, holidays, events, product affinity), and accuracy gains from outlier and anomaly handling. Refined forecasts weekly with buying offices, benchmarked new-product launches against analogous historical items. - Forecasting for short life-cycle and viral products. When a catalog includes items that live 2 weeks to 3 months, no single model covers slow movers, steady sellers, and viral hits. I build per-tier model strategies so each segment gets a model that fits it. - Forecasting for brand-new SKUs with no history. Bayesian methods that produce a usable forecast on day one and tighten as data arrives, with uncertainty surfaced to planners so they can size buffer stock to confidence rather than guess. - Anomaly detection that explains itself. Models that flag outliers, attribute likely causes, and recommend actions, with what-if dashboards for ROI simulation before committing. - Promotional uplift you can actually trust. Sensitivity analysis on noisy, contextual data so buying teams stop guessing the revenue impact of a promo before they run it. - Cloud data engineering and serverless systems. AWS Glue ETL pipelines for banking data, multi-account AWS environments via Control Tower and Organizations, and a fully serverless AWS product (Lambda, Amplify, DynamoDB, Cognito, SES, SNS, CloudWatch) I built end-to-end as co-founder. Plus Azure data engineering on retail platforms (Databricks, Delta Lake, Data Factory, App Service). Recent work: - Long-horizon retail demand forecasting (full-time, retail consultancy): sales forecasting solution for established products on Oracle RDF CS, with outlier handling for accuracy gains. Refined weekly with buying offices, benchmarked new-product launches against analogous historical items. - Fast-fashion e-commerce demand forecasting (Upwork client): product life cycles from 2 weeks to 3 months with wildly varying demand patterns. Delivered multiple model strategies tailored to product tier (two-stage classification then regression for sparse SKUs, Tweedie for steady movers, momentum-aware models for viral hits), with rolling-origin backtests and clean Python that dropped into their pipeline. - AWS infrastructure for ATM banking data (consulting mission): designed multi-account AWS environments with Control Tower and Organizations, automated VPC, IAM, RDS, S3 deployment via CloudFormation, built Glue ETL pipelines integrating with REST APIs. - AI-powered LinkedIn automation product (co-founder, technical builder), now live and in early customer rollout. Built the entire serverless AWS stack myself (Lambda, Amplify, DynamoDB, Cognito, IAM, SES, SNS, CloudWatch) plus the AI / LLM workflows for content generation and outreach. - Bayesian forecasting platform for short life-cycle products (current, full-time): cold-start forecasting in production on Azure App Service, Databricks, Scala backend. - Promotion uplift platform on Azure: full pipeline from Data Factory ingestion through PySpark on Databricks to a planner-facing front-end. - Shrink detection plus mitigation: end-to-end ML system enriched with scraped demographic and crime data, with an interactive what-if dashboard for prevention-action ROI. How I work: end-to-end ownership, business question first, clean documented Python that plugs into your existing pipeline, honest backtests, uncertainty surfaced rather than hidden. Stack: Python (Pandas, NumPy, Polars, Scikit-learn, Statsmodels, Prophet), PySpark, SQL, Scala, XGBoost, LightGBM. AWS (Glue, Lambda, S3, RDS, IAM, CloudFormation, Control Tower, Organizations, Amplify, DynamoDB, Cognito, SES, SNS, CloudWatch). Azure (Databricks, Delta Lake, ADLS, Data Factory, App Service). Project-based or longer engagements both fine.