Skills in Action

Nkosi Felix can build controlled ingestion workflows.

Demonstrated by: NBCU/Universal Creative document-ingestion harness (Graph/SharePoint delta-sync, validation routing).

Limits: Does not expose employer-confidential data or claim unsupported production scope.

Nkosi Felix can build Kafka-to-PostgreSQL pipeline patterns.

Demonstrated by: idempotent Kafka-to-PostgreSQL service with replay handling and DTO/schema validation at 4C.

Limits: Uses synthetic data, not the original employer's proprietary pipeline.

Nkosi Felix can add schema/DTO validation and failure routing to ingestion pipelines.

Demonstrated by: ACCEPT/QUARANTINE/SYSTEM-failure routing (NBCU), DTO validation (4C), Field Change Order Evidence Tracker lab.

Limits: Labs are synthetic; no real employer or customer data.

Nkosi Felix can surface observability and degraded modes in data systems.

Demonstrated by: Grafana/Prometheus/Loki dashboards at 4C (~30% fewer debugging cycles); this site's own fail-closed degradation.

Limits: The ~30% figure is the only allowed metric for that role.

Nkosi Felix can use GitLab CI/CD and deterministic test gates.

Demonstrated by: contract-test gates at 4C (JUnit 5, Mockito); Acceptance Harness (7/7); Air-Gap Ingest (26/26).

Limits: Test counts are from this project's own CI output, not a third-party audit.

Nkosi Felix can design AI assistants that only answer from published sources.

Demonstrated by: this site's own chat (see Hiring Manager Proof).

Limits: Does not claim unrestricted chatbot intelligence or private data access.

Nkosi Felix can use deterministic routing for public facts.

Demonstrated by: a fixed-rule chat router answering identity/employer/feature questions with zero LLM calls.

Limits: Covers a bounded set of public questions; unmatched ones fall to a provider-assisted or fail-closed path.

Nkosi Felix can use LLM providers behind a verification layer, not as a replacement for real sources.

Demonstrated by: live, provider-assisted answers, gated by evidence/safety/confidence checks before reaching a client.

Limits: Provider/model identity is never exposed; used only when selectable and healthy.

Nkosi Felix can build AI-assisted career product workflows.

Demonstrated by: GetBetterAI.com (see the case study for full detail).

Limits: Public product surface only; no private leads, payments, coaching records, or customer data shown.

Nkosi Felix can reject unsupported claims safely.

Demonstrated by: a fail-closed verification step, shown by the unsupported-claim rejection example on the homepage.

Limits: Rejection is deterministic and policy-driven, not a guess.

Nkosi Felix can support Resume Match / JD Fit as chatbot capabilities.

Demonstrated by: deterministic (no-LLM) resume-match and JD-fit-score logic in the chat assistant.

Limits: Does not rewrite the resume, guarantee interviews or offers, or store pasted job descriptions by default.

Nkosi Felix can design SharePoint/Microsoft Graph-style intake patterns.

Demonstrated by: document-ingestion harness and checkpointed Graph/SharePoint delta-sync at NBCU.

Limits: Does not expose private employer repositories or confidential data.

Nkosi Felix can use checkpointing, delta-sync, manifests, and provenance to make sync outcomes reviewable.

Demonstrated by: eTag/content-hash delta-sync, atomic landing, and run-evidence capture.

Limits: Described at the pattern level, not the original employer's internal repository.

Nkosi Felix can route validation outcomes into reviewable handoffs.

Demonstrated by: ACCEPT/QUARANTINE/SYSTEM-failure routing with structured reasons and ambiguity holds.

Limits: No real employer data; the lab uses synthetic change-order records.

Category: Health

Nkosi Felix can build synthetic-data health-intake labs.

Limits: Not a HIPAA-compliance claim.

Category: Energy

Nkosi Felix can build synthetic grid/telemetry quality monitors.

Category: Construction / Infrastructure

Nkosi Felix can build field change-order evidence trackers.

Category: Data / AI

Nkosi Felix can build evidence-grounded RAG evaluators.

Limits: Mirrors this site's own chat, not a separate production system.

Nkosi Felix can build deterministic validation harnesses.

Demonstrated by: Acceptance Harness — pure functions, no I/O, typed outcomes, 7/7 JUnit tests.

Nkosi Felix can build idempotent ingest pipelines with full observability.

Demonstrated by: Air-Gap Messaging Ingest & Observability — Kafka-to-Postgres, duplicate handling, Grafana/Prometheus/Loki, 26/26 tests.

Nkosi Felix can build clean-room commerce data lakehouses.

Demonstrated by: ShopLabs Lakehouse Lab — clean-room Python/DuckDB, synthetic Faker-generated order data, Parquet export.

Limits: No real customer, order, or payment data.

Nkosi Felix can build transparent resume/JD scoring services.

Demonstrated by: CareerAssistAI Scoring Pipeline — clean-room FastAPI service, synthetic tokenized keyword matching, transparent rationale.

Limits: No real recruiter or tracker data.

Nkosi Felix can build bilingual (EN/ES) site surfaces without duplicating content systems.

Demonstrated by: a dictionary-swap language toggle with localStorage persistence, applied across every public page.

Limits: Does not claim native Spanish fluency or a professional translation certification.

Nkosi Felix can expose machine-readable facts for AI agents and humans.

Demonstrated by: ai-index.json, llms.txt, sitemap.xml, claims-register.json, proof-pack.json, and a checksummed evidence manifest.

Limits: Machine-readable files describe public facts only; explicit "do not infer" sections guard against overreach.