Portfolio - Esperance Brooks

Security

Least privilege, isolation, threat surfaces

Cost

Predictable spend & pay-per-use where sensible

Performance

Latency, scalability patterns, efficiency

Maintainability

Clear boundaries & evolution paths

Each project below includes problem framing, architecture intent, service choices, and what I'd refine next.

Applied AI / Cloud Integration

Voice Command Event Processing Architecture

Prototype design for a lightweight speech-to-action system that converts spoken commands into secure, auditable operational events. Focus: modularity, isolation and event-driven extensibility.

Ingestion: Browser / device captures audio → uploads to pre-signed S3 URL.
Transcription Layer: AWS Transcribe triggered → JSON transcript stored.
Intent Extraction: Lambda (Python) applies lightweight parsing / optional Bedrock model call.
Event Bus: Clean intent published to EventBridge with metadata + roles claim.
Action Handlers: Downstream Lambda targets (e.g., ticket creation, simple automation).
Observability: Centralized structured logs + CloudWatch metrics for latency per stage.

Pattern Event-driven

Auth Cognito + IAM

Core Stack S3 · Transcribe · Lambda · EventBridge

Next iteration: Swap naive parsing for fine-tuned intent model; add replay protection & per-intent execution budget limits.

Diagram placeholder: sequence from capture → transcription → intent → event bus → handlers.

(Optional) GitHub Link

Accessible Learning / Low Bandwidth

Youth Learning Space (Africa)

Concept for a lightweight, offline-friendly learning hub designed for communities with intermittent connectivity. Emphasis on caching, progressive enhancement, and low operational overhead.

Delivery Model: Static site + API augmentation only when online.
Data Sync: Service Worker manages lesson bundle updates; conflict-light append-only progress logs.
Content Packaging: Versioned JSON modules + optional media compressed with adaptive quality.
Identity: Cognito pooled credentials or local-first guest mode with later binding.
Equity Focus: Minimal dependencies, energy-aware design, inclusive UI color contrast.

Stack: S3 · CloudFront · DynamoDB (progress) · Lambda · Cognito Pattern: Offline-first via Service Worker Focus: Resilience & accessibility

Next iteration: Add adaptive content prefetch heuristics + lightweight analytics respecting privacy.

Applied AI / Practical Advisory

Garden Design Advice App

A concept tool that guides homeowners on plant selection & layout using simple image inputs + contextual prompts. Prioritizes cost control over heavy ML training by leveraging managed AI services.

Input Flow: User photo(s) + basic constraints (sunlight, soil type) captured via form.
Processing: Lightweight image labeling (Rekognition) → enriched prompt to generative model for layout & species suggestions.
Recommendation Layer: Lambda builds structured plan (plant list, spacing, care schedule).
Delivery: Static front-end + on-demand generation; caching common climate/zone responses.
Governance: Guardrails for safe model outputs + transparent advisory disclaimer.

Stack: S3 · Lambda · Rekognition · Bedrock (model) Pattern: On-demand inference Focus: Cost vs utility balance

Next iteration: Add regional planting calendars + minimal progressive web app mode.

Core Domains & Tools

Cloud Architecture

AWS service selection, event-driven and serverless composition.

Security & IAM

Principle of least privilege, boundary design, auditability.

Performance Thinking

Latency awareness, decomposition, cost-performance trade-offs.

Applied AI

Pragmatic integration of managed ML + structured prompts.

Data Handling

Lightweight pipelines, storage patterns, integrity focus.

Iteration & Learning

Intentional reflection + next-step clarity per project.

Project Architecture Showcase