Spintel Syndication Engine
Radio play logs do not record which spins came from a nationally syndicated show and which the local station picked on its own. That gap is what lets a record label inflate a song's chart position by buying syndication. This is the detection engine that closes it. An 8-stage Python pipeline fingerprints song sequences across hundreds of stations, aligns the matches even when songs were added or dropped, clusters them into named shows, and tags every individual spin as syndicated or organic. The chart engine then publishes two rankings side by side: the public one and a stripped version showing real popularity. It runs daily on AWS as a serverless deployment, resolved 168 stations into 18 distinct shows, and replaced an analyst who tracked four to five shows by hand every morning.
0
45
Fannie Mae ETL Automation
Enterprise financial data moving through AWS at volume, with ETL that had become slow and hard to trace. The work automated those workflows across Glue, Lambda, S3, EMR and Athena, rebuilt the Python applications running them, and tuned SQL and PostgreSQL so processing held steady under load. Operational performance improved 30%. Every process step was documented, which mattered day to day as much as the speed gain, since recurring issues could then be resolved without re-tracing the whole pipeline. The result is a dependable data foundation that downstream reporting and analytics read from directly.
0
20
MAPLE / Parameter Tracking Tool
The tool answers one question for program managers: will this Key Performance Parameter reach its threshold before the program ends, and how confident can anyone be in that answer. It is a Windows desktop application that takes measured KPP data, fits seven regression families, and plots the forecast against threshold, objective and critical-path reference lines with automatic crossing detection. The MAPLE engine adds a maturity curve that keeps early-program forecasts conservative and a likelihood score reporting the probability of crossing month by month. V9 introduced a machine-learning calibration layer: trained coefficients for the likelihood formula, and a learned ensemble blend across all seven regression families. Both use classical optimization rather than neural networks, so the numbers stay defensible in governance review. It began as a demo BETA and grew across multiple major versions into a full forecasting engine.
0
36
A SaaS product with friction at two points in the funnel: account creation and everything after checkout. On the front end of that funnel, third-party authentication built with Python, Django and OAuth2 let people sign in with Google or Facebook instead of filling out a form, and signups rose 25%. On the back end, AWS-powered upsell automation lifted post-purchase conversions. The engagement also covered an AWS server migration and a structured onboarding process that tightened client communication across the software lifecycle.
0
58