Not project summaries — architectural solutions to specific problems like capture vs finalization and CQRS separation.
From Kayra Export Performance treated as separate from product UX
A 13-part engineering series on treating loading, responsiveness, and stability as product features—and protecting user-perceived performance under marketplace scale and AI workloads.
Start the 13-part performance series From Kayra Export Capture is easy; finalization is hard
How I closed the gap between capture and complete in a checkout payment engine with outbox, inbox, reconciliation, and effectively-once defenses.
Read the 22-part technical case study From Kayra Export When a shared CRUD model breaks under growth
Command/query separation, outbox, and production recovery when orders, stock, and reporting can no longer share one model.
Start the CQRS series
From ABC Logistics One panel does not scale; independent deploys do
How I split a logistics ops surface with Module Federation into a shell, auth remote, and feature remotes to shrink blast radius.
Read the 10-part MFE (Module Federation) case study From ABC Logistics Is the vehicle waiting — or is the record?
How I built waiting truth on the map from a telematics feed and configurable circle geofences (km_metric).
Read the 8-part geofence case study From ABC Logistics Partner capacity is not CRUD (Create, Read, Update, Delete)
How I built a ferry booking desk with ticket lifecycles, discount workflows, and a booking remote that embeds in a shell and runs standalone.
Read the 8-part booking case study From ABC Logistics Map, CMS (Content Management System), and notifications share one ops truth
How I kept map-first expeditions, CMS-like content, push admin, and an ERP edge on a deliberate control plane beside the MFE (Module Federation) estate.
Read the 8-part control plane case study From ABC Logistics ERP (Enterprise Resource Planning) MySQL is not an analytics store
How I cleansed internal-ERP MySQL into PostGIS, used MongoDB for telemetry and Elasticsearch for ops search across multi-node clusters, and served clean reads to frontends.
Read the 10-part data platform case study From ABC Logistics Recommended route ≠ driver preference
How I used EDA (Exploratory Data Analysis), scikit-learn/PyTorch, OSRM (Open Source Routing Machine), and foundation-model APIs to measure wait hotspots, km extension, fuel waste, and delay — then served them to the map UI.
Read the 10-part route intelligence case study From ABC Logistics A notebook is not an API; pipelines need security
How I bound a Django analysis API to MFE (Module Federation)/control-plane contracts, shipped via self-hosted GitLab CE → EC2, and embedded continuous ProjectDiscovery-style scanning in delivery.
Read the 8-part API & delivery case study
From Bare Crypto A landing funnel is not a mainnet wallet path
CBD All-Stars and Confident Gorillas: WIP whitelist, IPFS (InterPlanetary File System) metadata, placeholder→reveal, and Web3Modal mainnet hardening across four parts.
Read the 4-part mint & mainnet case study From Bare Crypto Reserve, auction, and emissions are not one contract
Remix + OpenZeppelin BareNFT, BareNFTReserve (randomBuy), BareNFTAuction, and BareToken emissions — with weak RNG and emergency-power trade-offs — across five parts.
Read the 5-part Solidity protocol case study From Bare Crypto Scanning totalSupply is not an OpenSea index
Bare Crypto marketplace SPA: OpenSea-like IA, N RPC (Remote Procedure Call) listing discovery, and Rinkeby vs mainnet client split — three parts.
Read the 3-part marketplace UI case study From Bare Crypto A marketing SPA is not a Chrome extension
Trait generator → mint metadata pipeline and an honest Bare Wallet landing scope (funnel SPA, not MV3) — two parts.
Read the 2-part product surface case study
