The Hidden Cost of AI Context: Kamel Charaf, Compresr (EP#579)
September 17, 2026
Co-produced by
Co-produced by
04:14 – Why Context Management Matters
10:34 – How Compresr Works
20:07 – Getting Into Y Combinator
36:03 – What People Get Wrong About AI
Kamel Charaf is the Co-Founder and CPO of Compresr, a company helping AI teams reduce the amount of context they send to large language models. Instead of building another AI application, Compresr sits one layer lower: it decides what information the model actually needs before the request reaches it.
In this episode, Kamel explains why context management is becoming a real infrastructure problem for AI companies. As more teams use agents, company brains and large document workflows, the amount of text sent to models keeps growing. That can make AI slower, more expensive and less reliable.
Kamel also talks about building Compresr out of the EPFL orbit, joining Y Combinator, moving between Switzerland and Silicon Valley, and learning to shift from a research mindset to a more commercial one. The conversation gets into token costs, enterprise pilots, on-premise deployment, Swiss talent, YC’s “bro style” advice and why the next layer of AI may be less flashy than the apps everyone talks about.
The cover portrait was edited by www.smartportrait.io.
Don’t forget to give us a follow on Instagram, Linkedin, TikTok, and Youtube so you can always stay up to date with our latest initiatives. That way, there’s no excuse for missing out on live shows, weekly giveaways or founders' dinners.
%2520(5).avif)
