12mo roadmap · 5 milestones
12-Month Advanced AI Infrastructure
The long arc. A year of compounding depth across search, vector databases, AI systems, backend, storage, and system design — measured in shipped artifacts and project features.
Outcome: Reach the depth to design and reason about systems like Turbopuffer, and ship them through HighSignal and Codevetter.
How to work this roadmap
Treat each milestone as a claim that must be supported by evidence. Before reading, write what you think the mechanism does and where it will fail. After studying the linked concepts, run the drills without copying an answer, preserve the output, and revise the explanation. Move forward when you can connect the milestone goal to a working implementation, benchmark, architecture decision, or reviewable design artifact.
The 12mo horizon is a sequencing aid, not a completion badge. Spend more time where your prediction and the observed behavior disagree. Keep a short decision log containing the mechanism selected, alternatives rejected, expected failure mode, measurement used, and remaining uncertainty. Review that log with FSRS prompts so the roadmap produces durable system judgment rather than a temporary tour of terminology.
At the end, explain Reach the depth to design and reason about systems like Turbopuffer, and ship them through HighSignal and Codevetter. from first principles to a reader outside the domain. A strong explanation should survive follow-up questions about correctness, cost, latency, resource use, security, recovery, and operational visibility. If it cannot, return to the milestone that contains the missing mechanism and build a smaller falsifiable example.
Milestones
Milestone 1
Prerequisites — unlock the rest of this path
Concepts the later milestones depend on. Without these the planner cannot serve the rest of this roadmap.
Concepts
Milestone 2
Q1 — Retrieval and vector foundations
Complete the 90-day AI search/infra roadmap.
Concepts
Build evidence
- Hybrid search v0 in HighSignal — Combine BM25 and vector retrieval with reciprocal rank fusion.
- Brute-force vector index — An exact nearest-neighbour index — the recall=1.0 baseline.
- RAG pipeline v0 — An end-to-end retrieve → pack → generate pipeline grounded in a document corpus.
Milestone 3
Q2 — Production backend depth
API platform strength: keys, limits, idempotency, queues, observability.
Concepts
Build evidence
- Reusable API key system — Issue, hash, scope, and rotate API keys for machine clients.
- Reusable rate limiter — A token-bucket rate limiter usable as middleware.
- Job queue with retries and DLQ — A background job queue with backoff, jitter, and a dead-letter queue.
- Structured logging package — A structured logger with a request correlation id.
Milestone 4
Q3 — Storage systems
Build the storage foundations behind a Turbopuffer-class system.
Concepts
Build evidence
- Toy write-ahead log — An append-only WAL with crash recovery.
- Toy LSM tree — A minimal LSM storage engine: memtable, SSTables, compaction.
- Object-storage-backed index — Store index segments in object storage with a hot in-memory cache.
Milestone 5
Q4 — System design and synthesis
Turn the depth into architecture-level design and a real prototype.
Concepts
Build evidence
- Search platform design doc — An architecture doc for a search platform: ingestion, indexing, query, ranking.
- RAG system design doc — An architecture doc for a production RAG system with an eval gate.
- Model router — Route each request to the cheapest model that passes evals for that request class.