Enterprise Order & Payments Platform
A modular .NET payments reference with idempotency keys, a transactional outbox and signed webhooks. Explores provider-timeout recovery, refunds and settlement reconciliation against simulated integrations.
The engineering library
Fifty independent projects spanning systems engineering, modernization and AI applications. Each has its own implementation, documentation and explicit scope. Three include a curated, in-depth case study.
Use a broader term, choose another language or clear the filters to see the full library.
Backend and platform reference systems: payments, ledgers, scheduling, identity and data platforms built as self-directed studies.
A modular .NET payments reference with idempotency keys, a transactional outbox and signed webhooks. Explores provider-timeout recovery, refunds and settlement reconciliation against simulated integrations.
A paired legacy-style claims application and modern .NET target. Characterization tests, an anti-corruption layer and a data importer make compatibility and phased migration decisions inspectable.
A .NET knowledge-assistant reference combining permission-aware hybrid retrieval, citation spans and grounding checks. Deterministic local embeddings support repeatable retrieval and access-control evaluation.
A multi-tenant .NET platform with isolation at request, application and persistence boundaries. Includes authorization, entitlements, quotas, feature rollout and a simulated billing lifecycle.
A rules-driven settlement reconciler supporting fee-adjusted and many-to-one matching. Unresolved records enter a four-eyes review queue, with repeatable runs and an auditable exception lifecycle.
A .NET diagnostics lab that reproduces failure modes such as pool exhaustion, thread starvation and retry storms. Toggleable faults connect application behavior to measurements and corrective changes.
Separate customer, partner and administrator API surfaces with differentiated trust policies. Examines token rotation, authorization, signed webhooks and audit trails within a controlled .NET reference.
An event-time fleet pipeline handling late, duplicated and out-of-order telemetry. Partitioned processing, deduplication and replay support geofencing, hysteresis and ETA calculations on simulated GPS data.
An accounts-payable document pipeline with field extraction, confidence thresholds and human review. A seeded corpus makes extraction quality and straight-through-processing trade-offs measurable.
A C# MQTT codec, broker and edge gateway exploring telemetry delivery through connectivity loss. Store-and-forward behavior, device state and anomaly rules are evaluated on a simulated sensor estate.
A multi-channel notification reference with tenant scheduling, quiet hours, delivery receipts and replay. Simulated unreliable providers expose circuit-breaking and failover behavior.
A configuration-driven integration hub with versioned connectors and a restricted mapping evaluator. Simulated systems exercise resumability, contract drift, dead-letter replay and boundary controls.
An append-only, double-entry ledger using integer money and explicit concurrency controls. Explores posting invariants, holds, reversals and FX workflows under controlled concurrent execution.
A versioned loan-decision workflow combining declarative rules, explainable scorecards and simulated external checks. Decision traces and what-if runs make underwriting changes reviewable.
A transaction-risk pipeline with partitioned processing, windowed features and versioned rules. Explores velocity, travel and device signals alongside latency budgets and analyst feedback.
A subscription-billing reference covering pricing versions, proration, late usage events and idempotent invoices. Explicit dunning and suspension states make lifecycle behavior easier to inspect.
A multi-node job scheduler using leases and fencing tokens to coordinate workers. Covers stalled-worker recovery, cron and DST handling, and dependency-aware execution in a controlled environment.
A feature-flag control plane and .NET SDK with local evaluation, stable percentage bucketing and streamed updates. Explores offline configuration fallback and visibility into stale flags.
A tenant-aware audit store with hash chaining, Merkle checkpoints and inclusion proofs. Examines tamper detection alongside retention, legal holds, redaction and auditing of reads.
A resumable credential-rotation workflow spanning consumer notification, verification, promotion and rollback. Models envelope encryption and controlled emergency access without claiming a deployed secrets service.
A collaborative-editing backend using an RGA text CRDT and versioned structured fields. Randomized checks explore convergence, causality buffering, reconnect behavior and comment-anchor rebasing.
A C# search reference with a positional index, BM25, vector search and rank fusion. Relevance evaluation connects indexing and ranking decisions to measured retrieval behavior.
A clinic-workflow reference covering scheduling constraints, concurrent bookings and time zones. Amendment-only notes and care-relationship authorization make data and access boundaries explicit.
An identity-governance model with role inheritance, policy evaluation and lifecycle automation. Synthetic identities exercise separation of duties, temporary elevation and access-review workflows.
A lakehouse reference with transactional table snapshots, change-data ingestion and history-aware transformations. Data-quality gates, backfills and column lineage connect ingestion to serving correctness.
An SRE reference deriving service indicators and multi-window burn-rate alerts from controlled telemetry. Error-budget policy connects reliability measurements to deployment decisions.
A load-testing toolkit comparing arrival-rate and virtual-user models. Examines tail latency, coordinated omission, saturation and soak behavior instead of relying on average response times.
A deployment reference combining migrations, health checks, configuration and graceful draining with infrastructure templates. Documents staged rollout and rollback mechanics without claiming a live cloud deployment.
An agent-workflow reference with a registered tool allowlist, validated arguments, approval steps and durable execution. Budget controls and replayable checks make the authority boundary explicit.
A FinOps reference reconciling synthetic cloud spend to team allocations. Examines seasonal anomalies, forecast behavior and modeled savings rather than presenting simulated results as customer savings.
Legacy-to-modern migration mechanics: interop boundaries, strangler routing, saga verification and cutover tooling, studied in controlled environments.
A .NET IL analyzer examining assembly dependencies, compatibility concerns and migration ordering. A generated assembly estate makes static-analysis coverage and cycle-breaking decisions reproducible.
A Go shadow proxy combining request mirroring, semantic JSON comparison and controlled cutover states. A labeled response corpus exposes where comparison rules can misclassify migration differences.
A Java migration verifier comparing H2 and SQLite records across decimal, time, Unicode and null-handling differences. Controlled fixtures expose where normalization can conceal data corruption.
A C# state-space explorer for extracted sagas. Examines failure interleavings, compensation reachability and recovery properties within an explicit bounded model.
A controlled C++/.NET interoperability study of boundary contracts, resource ownership and numerical compatibility. Separates interface hardening from compiler effects around unchanged engine source.
A .NET authentication-coexistence lab covering Forms cookies, WS-Federation and OIDC. Examines claims mapping, rehash-on-login upgrades, downgrade controls and rollback assumptions.
A seeded Rust simulator for a quorum-replicated register. Injects partitions, clock skew and network reordering so failure histories can be replayed and consistency assumptions examined.
A Python migration planner combining dependencies, capacity constraints, cost models and Monte Carlo scenarios. Completion estimates describe modeled uncertainty, not delivery commitments.
A PostgreSQL DDL linter and lock-queue simulator in Go. Explores migration rewrites and adaptive backfill against lock-duration budgets in controlled scenarios.
A COBOL record decoder and controlled batch-replacement comparison covering EBCDIC, packed decimal and variable-length layouts, with byte-equivalence checks on generated fixtures.
Applied AI systems work: retrieval evaluation, agent runtimes, evaluation harnesses and observability, measured with offline or simulated data.
An offline LLM-evaluation harness with paired statistical comparisons, judge-agreement diagnostics and power analysis. Distinguishes observed score changes from evidence sufficient for a regression gate.
A deterministic Go simulator examining semantic-cache false hits, precision/cost trade-offs and request coalescing. Results describe modeled traffic, not live provider latency or billing.
An offline lab attributing evidence loss to chunking, retrieval or ranking. Exact spans in a synthetic corpus support controlled comparisons, recorded predictions and explicit statistical limits.
A controlled prompt-injection evaluation separating capability and egress controls from heuristic defenses. Attack and benign corpora expose where protection depends on execution assumptions.
A journaled agent runtime exploring recovery and external-side-effect duplication through crash injection. Execution guarantees are evaluated within the documented gateway and idempotency contract.
A C++ compiler for a bounded JSON Schema subset, with token masks and Python bindings. Measures validity, masking cost and distribution bias using a finite-state language-model study.
A discrete-event inference-gateway simulator comparing batching, admission control and tenant fairness. Queueing analysis explains capacity and latency behavior under modeled traffic.
A Java study comparing graph traversal with retrieval-induced subgraphs over generated ownership filings. Evaluates multi-hop queries, entity-resolution errors and provenance failure modes.
A Go evaluation of routing, cascade and difficulty-classifier policies against cost/quality baselines. Separates routing gains from improvements available through fixed model selection.
A simulated answer-quality monitoring study using deterministic character 4-gram features. Compares detection, false positives, delay and modeled cost without live embeddings or model calls.