Software Engineering Daily

Small AI Models with Yoeven Khemlani

91 snips
Jul 24, 2025
Yoeven Khemlani, the founder of JigsawStack, shares his journey from game development to innovating small AI models for backend applications. He and tech expert Gregor Vand dive into the advantages of these models, highlighting their speed and cost efficiency. They discuss advancements in web scraping, data translation, and the importance of fine-tuning for consistent outputs. Yoeven also touches on community engagement through hackathons and the challenges of solo entrepreneurship, emphasizing the need for strategic market moves.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
ADVICE

Stabilize Outputs via Prompt Baselines

  • To maintain consistent outputs, Jigsaw Stack stores prompt outputs over multiple runs to establish baselines and accuracy scores.
  • Over time, prompt execution narrows to the single best performing model, reducing variability and preventing breaking changes.
INSIGHT

Small Models Enable Broad Deployment

  • Jigsaw Stack targets GPU-poor environments by optimizing for smaller, cost-efficient GPUs and deployable models.
  • Owning proprietary small models allows easier deployment anywhere, solving enterprise distribution and hosting challenges seen with large LLM providers.
ADVICE

Easy and Intuitive Developer Experience

  • Getting started with Jigsaw Stack is simple: install the NPM or Python package and use the fully typed API with descriptive fields.
  • The API structure is intuitive and consistent, designed for ease of use without heavy documentation.
Get the Snipd Podcast app to discover more snips from this episode
Get the app