ShapedQL

What it does

ShapedQL is a SQL engine designed to handle multi-stage ranking and retrieval-augmented generation (RAG) workflows. It aims to consolidate several separate infrastructure layers—vector databases, feature stores, inference services, and custom application logic—into a single system that can execute ranking queries through SQL.

Who it is for

The tool targets teams building personalized feeds, recommendation systems, or RAG applications that require both initial retrieval and ranking stages. The founder notes this includes products like social feeds and long-term memory systems in AI applications. Teams currently stitching together multiple infrastructure components would be the primary audience.

Pricing

The site does not show prices.

How it stands out

ShapedQL addresses what the founder identifies as a specific gap: vector databases have solved retrieval at scale (finding large candidate sets), but ranking—filtering down to the best items—remains fragmented across multiple tools. The engine consolidates vector retrieval, feature storage, model inference, and ranking logic into SQL queries, potentially eliminating thousands of lines of orchestration code. This reduces operational complexity and the number of systems engineers need to maintain.

What a founder should check

Any competitor should investigate three areas. First, understand how entrenched incumbents are: Pinecone, Milvus, Redis, and inference platforms like Replicate or Hugging Face each serve specific needs, and teams may have significant switching costs if already invested in these systems. A single replacement product must be dramatically better at all stages, not just comparable.

Second, verify the actual pain point severity. Interview teams currently building these systems to understand whether the integration burden is severe enough to justify abandoning established, well-understood components. Some teams may prefer modularity and the ability to swap components independently over having everything in one system.

Third, assess the technical moat. SQL engines for specialized workloads face competition from traditional databases adding vector support (Postgres, MySQL) and vector platforms building ranking features. Determine whether ShapedQL's approach meaningfully outperforms these expanding alternatives, or whether it simply repackages them. The founder's background in ranking (Instagram Reels) and machine learning infrastructure is a credential, but the actual competitive advantage in the market needs validation against both specialized and generalist solutions.