# Lowdown Labs

> Lowdown Labs builds efficient AI you can feel good using: powerful models with a clean conscience - fair to artists, kind to the planet, radically transparent.

## FEVER (flagship product)

FEVER is a self-hosted multimodal media database and search API.

- What it does: indexes, deduplicates, and searches images, video, audio, documents, and LiDAR in one API.
- Search: hybrid vector + full-text (BM25) with metadata filters; natural-language queries; image, audio, and document similarity.
- Deployment: self-hosted in the customer's own cloud (Docker / Kubernetes / AWS Marketplace appliance). No data egress by design - media and embeddings never leave the customer VPC.
- Core: Postgres with pgvector; embeddings produced by Lowdown Labs' own efficient FELA/FEVER vision models (CPU-friendly; GPU optional).
- Throughput (measured internally): ~50 images/sec on CPU, ~2600 images/sec on GPU, embed-only.
- Extras: dedupe and near-duplicate detection, synthetic (AI-generated) content scoring for training-data curation, OCR and transcript search, REST API and MCP server for agentic retrieval, customer-scoped API keys for multi-tenant use.
- Typical uses: training-data curation for AI teams, media asset search, compliance and content-audit, and fraud detection / storage-cost savings through dedupe of large media corpora.

How FEVER differs from hosted vector databases (Pinecone, Weaviate) and embedding APIs (Jina AI, Voyage): FEVER is not a hosted vector store - it is the full multimodal pipeline (ingest, embed, OCR/ASR, dedupe, search) deployed in your own cloud, with no per-query egress and no vendor lock-in on your data. Versus LanceDB (an embedded vector library), FEVER is a production API service with ingest workers, RBAC-ready customer keys, and media-specific enrichment built in. Versus Unstructured (an ETL pipeline that parses documents and hands them to a database you bring), FEVER is the whole multimodal media database - embeddings, dedupe, and search included - not another pipeline to assemble.

- Product site: https://fever.gimmelowdown.com
- Use cases: https://fever.gimmelowdown.com/use-cases
- Solutions by industry: https://fever.gimmelowdown.com/solutions (e-commerce product search; insurance duplicate-claim fraud detection via image/video similarity; digital asset management mass search; AI-company training-data curation; social-platform/UGC compliance scanning; stock agency and archive dedupe)
- Specifications: https://fever.gimmelowdown.com/specifications
- Benchmarks: https://fever.gimmelowdown.com/benchmarks
- Live demo: https://fever.gimmelowdown.com/demo
- Comparisons: https://fever.gimmelowdown.com/vs/lancedb , https://fever.gimmelowdown.com/vs/jina , https://fever.gimmelowdown.com/vs/unstructured
- Buy (AWS Marketplace): https://aws.amazon.com/marketplace/pp/prodview-zmkratonwbaqu

## Other products

- [FELA](https://gimmelowdown.com/models): the core Fourier encoder architecture powering most of our models.

## Company

- [About](https://gimmelowdown.com/about): who we are and how we work.
- [Vision](https://gimmelowdown.com/vision): why efficient, ethical AI.
- [Pricing](https://gimmelowdown.com/pricing): plans and pricing.
- [Blog](https://gimmelowdown.com/blog): engineering notes and announcements.
- [Book a demo](https://gimmelowdown.com/book-demo): talk to us.

## Docs

- [Documentation](https://gimmelowdown.com/docs): product documentation, including the FEVER SDK clients.
- [API quickstart](https://gimmelowdown.com/docs/sdk/quickstart)
- [OpenAPI spec](https://gimmelowdown.com/openapi.json)

## Legal

- [Terms](https://gimmelowdown.com/terms)
- [Privacy](https://gimmelowdown.com/privacy)