compresr Analysis
What is compresr?
compresr builds context compression tools that shorten prompts and documents before they reach an LLM. Its query-aware models keep answer-bearing spans and drop irrelevant text, which it says preserves answers while lowering token usage. The product is available through SDKs, HTTP, and framework integrations, with on-prem deployment for regulated environments.
Location
San Francisco, United States
Employees
1-10
Founded
2026
Industry
Infrastructure
Business Model v4
Developer infrastructure
Delivery Model
Developer platform / API / infra
Functional Type
AI infra / research,Business Intelligence
YC Batch
W26
YC Partner
Jared Friedman
Product Features & Capabilities
- cmprsr-v1 model for context compression
- SDK for easy integration
- Pre-compressed knowledge bases
- Custom compression models for specific domains
- Query-specific compression capabilities
Use Cases
Compress context for LLM queries to save on token usage; Utilize pre-compressed knowledge for efficient data retrieval; Implement custom compression for finance and legal documents; Achieve extreme compression rates for specific queries; Reduce API costs for engineering teams
Who are the founders of compresr
KC
Kamel Charaf
Co-founder & COO/CPO
OG
Oussama Gabouj
Co-founder and CTO
BA
Berke Argın
Co-founder & CAIO
IZ
Ivan Zakazov
Doctoral Assistant (PhD) at EPFL
Product Description
Compresr specializes in context compression technology for large language model (LLM) agents, offering their proprietary model, cmprsr-v1. This model achieves up to 200x compression of context without loss of quality, making it highly effective for various applications, including finance, legal, and healthcare sectors. Key features of cmprsr-v1 include both coarse-grained and fine-grained compression methods, which allow users to reduce context to only the most relevant information. This technology is designed to improve cost efficiency, reduce latency, and enhance accuracy for AI applications, enabling users to cut token costs and build smarter AI solutions. Additionally, Compresr provides an SDK for easy integration and pre-compressed knowledge bases to facilitate implementation.
Other Considerations
Backed by Y Combinator; Achieves up to 64% cost reduction for API usage; Demonstrated accuracy improvement of +3% with 2X compression
See something that needs updating? Suggest edits to this profile.