HomeArchitecture sprint
Architecture Sprint - One Hard Decision in 1-2 Weeks
An architecture sprint is a fixed one- to two-week engagement ($2,500) where a senior engineer diagnoses one hard architectural decision - such as monolith vs. microservices, build vs. buy, database scaling, or AI pipeline design - and delivers a written recommendation with trade-offs, not a vague slide deck.
What is an architecture sprint?
An architecture sprint is a time-boxed engagement focused on exactly one decision that blocks your team, rather than open-ended consulting or a multi-month rewrite proposal.
You bring the problem (such as 'should we split the monolith before Series A?', 'which LLM routing layer do we need?', or 'can this Postgres schema survive 10× traffic?'). I bring structured analysis, options with trade-offs, and an executable recommendation.
Typical decisions I answer
Sprints work best when the question is specific enough to decide in two weeks. Vague requests are scoped down during the kickoff call.
- Monolith vs. services: when to split, what to extract first, and migration path.
- Build vs. buy: evaluate vendor vs. in-house for a specific capability (auth, search, AI inference, etc.).
- Data layer: Postgres scaling, read replicas, caching strategy, or when to add a queue.
- AI architecture: single-call vs. agent, model routing, eval strategy, and cost projection.
- Integration design: third-party API, webhook pipeline, or event-driven refactor scope.
How the sprint runs
Week 1: discovery call, repo and doc access, async questions. Week 2: analysis, optional live working session, written deliverable, and readout call.
- Day 1-2: Kickoff call plus ingest codebase, ADRs, and constraints (team size, timeline, budget).
- Day 3-7: Analysis, spike review, and optional 60-minute working session with the engineering lead.
- Day 8-10: Written decision doc and 30-minute readout with founder/CTO.
Deliverable format
The output is an Architecture Decision Record (ADR): context, options considered, trade-offs, recommendation, and immediate implementation steps your team can commit to Git.
Proof points
- Production AI systems at scale: media localization, dubbing pipelines, AI twins
- Stack: TypeScript, Next.js, Python, FastAPI, Postgres, LLMs, MCP
FAQ
How is this different from the monthly advisor retainer?
The sprint is fixed scope and price for one decision. The monthly advisor is ongoing judgment across many problems. Many founders start with a sprint on the hardest blocker, then move to retainer.
Do you implement the recommendation?
No, brain, not hands. The deliverable is the decision and implementation plan. Your engineers execute; I am available for a follow-up call if questions arise during rollout.
What if the problem is too big for two weeks?
We scope down to the decision that unlocks everything else, or I will say the engagement is wrong and recommend embedded weekly presence instead.
Can this cover AI/LLM architecture?
Yes — common sprint topics include model routing, agent vs. single-call design, eval coverage, and cost projections based on production patterns from my operating team.
Next step
30-minute discovery call: review fit and disqualify honestly. A one-page scope doc follows within a few days if we proceed.