Available for fractional & project-based work

CastleBytes — Software Architecture & Consulting

Hi, I'm Gus, a senior software architect who designs and ships SaaS platforms, APIs, data pipelines, and AI integrations (MCP). A hands-on builder with a track record in cloud (AWS/GCP), databases, and security & identity. I help founders and product teams go from roadmap to production — balancing speed, reliability, and security.

  • SaaS & Multi-tenant
  • APIs & GraphQL
  • PostgreSQL · SQLite · DuckDB
  • MCP adapters
  • RBAC · OAuth · OIDC
  • AWS · GCP

What I do

I'm a lean software engineering partner that crafts secure, cloud-native platforms—from sleek React dashboards to high-performance Go, TypeScript, and Python back ends. Bring a concept or a legacy tangle—I'll architect, implement, and fortify it end-to-end.

Architecture Sprints

Greenfield design or legacy untangling. From moduliths to microservices: domain modeling, tenancy strategy, storage layout, and service boundaries—with reference designs that spell out risks, trade-offs, and cost.

  • Roadmap → reference design
  • Build vs buy calls
  • Risk & cost trade-offs

API & Data Engineering

Design/implement REST, GraphQL, and gRPC with durable schemas and SDKs. Event-driven pipelines (Kafka/NATS) and columnar flows with DuckDB + Parquet for analytics-heavy apps and real-time UX.

  • Schema & migration plans
  • Performance, caching, observability
  • Lineage & reproducibility

AI Integrations (MCP)

LLM-powered copilots and workflow agents with Model Context Protocol adapters. Memory systems to reduce token load and latency; governance hooks keep data safe.

  • Adapter design & hardening
  • Data classification & PII handling
  • Audit & policy controls

Security, Identity & Compliance

Identity and access done right: OAuth2/OIDC, SAML, SCIM, JWT, and scoped tokens. Service-to-service authentication, secrets management, and policy-as-code. Full audit trails and data governance to meet compliance needs without slowing teams down.

  • RBAC/ABAC, tenant isolation, scoped tokens
  • Identity federation & access reviews
  • Audit trails, lineage & compliance hooks
  • Gateway hardening, rate-limiting & observability

Modern Front-end Engineering

High-performance interfaces with React, SolidJS, or Astro. Bespoke admin dashboards, service radars, and data-driven UIs with strong DX and testability.

  • Design systems & theming
  • Server-driven UI & SSR/ISR
  • Accessibility & performance budgets

Fractional CTO

Hands-on leadership for 1-2 days/week: hiring loops, standards, and unblockers that keep teams shipping. Architecture reviews, implementation sprints, and ongoing advisory.

  • Technical strategy
  • Team enablement & code reviews
  • Vendor & stack selection

Selected outcomes

A few representative projects—from concept to fortified platform.

AI platform: MCP adapters + memory

Integrated internal tools and data via Model Context Protocol. Built a PostgreSQL + DuckDB memory layer that cached embeddings and intermediate results—cutting agent token load by ~40% and p95 latency by ~30%—with audit trails and policy hooks for sensitive data.

  • PII classification & redaction pipelines
  • Governance & observability end-to-end

Multi-tenant SaaS control/data plane

Control plane with per-tenant SQLite (WAL/LiteFS) and analytical data plane with DuckDB + Parquet. Real-time SSE, governed endpoints, lineage, approvals, and RBAC with org/user grants.

  • ULID/UUIDv7 IDs & schema-versioned artifacts
  • Auditable, least-privilege defaults

Fintech API modernization

Re-platformed from legacy SOAP/monolith to REST/GraphQL with OAuth/OIDC, rate-limiting, and full observability. Developer onboarding dropped from weeks to days with clean versioning and SDKs.

  • Gateway hardening & threat modeling
  • Backward-compatible migration plan

Data pipelines & analytics UX

CSV/Parquet registry with schema inference, versioning, and lineage. Streamlined notebooks into governed, reusable endpoints powering dashboards and internal tools.

  • Columnar storage + vectorized execution
  • Developer experience: templates & scaffolds

Core stack & tools

Go · Python · TypeScript · React · PostgreSQL · DuckDB · SQLite · GraphQL · AWS · GCP

Also: Kafka · NATS · Redis · MongoDB · Terraform · Snowflake · Prometheus/Grafana · NodeJS · Parquet

How engagements work

  1. 01
    Discovery
    Goals, constraints, and success criteria.
  2. 02
    Architecture
    Reference design, risks, milestones.
  3. 03
    Build
    Hands-on implementation with PRs & reviews.
  4. 04
    Harden
    Security, observability, and load tests.
  5. 05
    Transfer
    Docs, handoff, and hiring support as needed.

Engagement models

  • Architecture Sprint (1-2 weeks)
  • Implementation Project (4-12 weeks)
  • Fractional CTO (1-2 days/week)

Deliverables you can expect

  • Reference architecture & decision log
  • Executable code & tests
  • Runbooks, IaC, and onboarding docs

Fit & focus

Best fit for seed → Series B startups and product teams who need to accelerate a roadmap, de-risk a build, or integrate AI safely.

About me

I'm Gus, a hands-on software architect with deep experience across Go, Python, and TypeScript/React. I've built multi-tenant SaaS platforms, real-time data products, and AI integrations using the Model Context Protocol (MCP). My approach blends pragmatic architecture with builder's speed, strong security defaults, and a bias for maintainability.

Previously: led platform and data initiatives in fintech, identity, and analytics. Comfortable pairing with teams, mentoring, and leaving systems better than I found them.

Highlights

  • Built MCP adapters + memory systems to cut agent token and latency costs.
  • Designed governed endpoints with lineage, approvals, and audit trails.
  • Implemented tenancy databases control planes with redundant data planes.
  • Secured APIs with OAuth/OIDC, RBAC/ABAC, and threat-modeled gateways.

Let's talk

Start with a short intro call. In 30 minutes we can align on goals, constraints, and where I can help fastest.

  • Quick context & goals
  • Feasibility & approach options (REST/GraphQL, data flows, AI/MCP fit)
  • Risks & constraints (security, compliance, timeline, budget)
  • Rough scope & timeline (sprint or project)
  • Immediate next steps (I'll follow up with a short plan/estimate)

No prep required—bring an idea, a repo, or a deck. If you'd like, share links when booking.