Jeevan Siddhabhaktula

Risk work fails in a predictable way: a threat gets named, everyone agrees it is serious, and nothing changes — because nobody wrote down which signals actually distinguish it from an ordinary bad day. So I built the layers in order.

1. A data model. An open taxonomy of freight fraud patterns, where every indicator is weighted, stage-tagged, and paired with the innocent explanation that mimics it.
2. An instrument on top of it. An assessment engine that turns what you can actually observe into a ranked, arguable finding — and refuses to call it a probability.
3. The same discipline elsewhere. A control room that derives which AI regulations a system attracts, and holds the evidence behind every control.

Every tool here is static, dependency-free and run entirely in your browser. Nothing you type is transmitted, logged or stored.

• No open findings across the 9 monitored repos — all clean.

• Tested against the deployed site, not localhost. Every tool is checked by headless assertions run against its public URL, including an independent recomputation of its own arithmetic. A build that only passes on my machine has not been tested.
• The data validates itself. Each repository ships a schema and a dependency-free validator that runs in CI and refuses a malformed model, so a bad edit fails loudly rather than rendering a plausible wrong answer.
• No invented numbers. The atlas has no geographic heat map, because credible cargo-crime incident data is paywalled and I will not fabricate hotspots to make a page look better. Where a figure is modelled rather than observed, it says so.
• Nothing employer-specific. No internal thresholds, no real carriers, no case data. The judgement is transferable; the confidential material stays where it belongs.

Risk Manager at Amazon since 2021, on a central risk team handling carrier fraud investigations across EU and NA lanes — missing trailers, pilferage, identity fraud, double brokering, GPS spoofing and internal collusion — working with business conduct, insurance and worldwide operational security. Lean Six Sigma Black Belt.

Two numbers from that work explain everything on this page. The first is a ~40% reduction in EU and NA audit false positives, achieved by redesigning audit logic and exclusion rules — which is why every pattern in the taxonomy above carries a false_positives block, and why the Atlas refuses to treat missing evidence as exculpatory. The second is >$15M in model-based prevention impact from pattern analysis and control reinforcement, alongside a risk-forecasting engine at roughly 95% accuracy and reporting automation that removed about 90% of the manual effort.

The recurring lesson is the one these tools encode: the expensive failures are almost never a missing signal. They are a signal nobody could separate from noise in time to act.

Certifications — Lean Six Sigma Black Belt · Lean Six Sigma Green Belt · ISO 28000 Supply Chain Security · ISO 9001 Internal Auditor · AWS Certified Solutions Architect · Cybersecurity Foundations, University of London · Power BI / Data Analytics

I also write. A Conversation With Existence is a philosophical memoir, and it is here for the same reason the code is: I would rather be judged on things I actually finished.