Skip to main content

SqueezeBits

Category: AI Infrastructure

A deep-tech AI infrastructure startup specializing in hardware-aware AI model compression and acceleration to enable efficient LLM deployment and on-device AI. SqueezeBits was founded in 2022. The company is led by Hyungjun Kim. Based in Seoul, South Korea. Team size: 10-50. Total funding raised: $2.6M. Latest round: Strategic investment (Undisclosed, Dec 2024). Key investors include Kakao Ventures, Samsung Next, Naver D2SF, POSCO Capital, POSTECH Holdings, Rebellions.

Founded
2022
Headquarters
Seoul, South Korea
Team size
10-50
Total funding
$2.6M

Value proposition

Reduces AI operational costs and latency by up to 4x while maintaining model accuracy, enabling massive LLMs to run on resource-constrained hardware and high-performance NPUs.

Products and solutions

OwLite: A SaaS-based AI quantization toolkit that automates model compression for various hardware targets (NVIDIA, Intel, Qualcomm)., Fits on Chips: An LLM serving optimization solution that identifies the most efficient hardware/software configuration for large-scale model deployment., Yetter: A high-performance generative AI API service designed for cost-effective and low-latency inference., Custom NPU Optimization: Specialized services for hardware-specific AI acceleration and performance benchmarking.

Unique value

Founded by a PhD-led team from POSTECH with expertise in both AI algorithms and hardware (NPU) design, allowing for 'hardware-aware optimization' that outperforms pure software approaches.

Target customer

AI model developers, enterprise software companies, NPU (Neural Processing Unit) manufacturers, and edge computing device providers.

Industries served

AI Infrastructure, Semiconductor (NPU/GPU), Mobile & IoT, Automotive (Autonomous Driving), Enterprise LLM Deployment

Technology advantage

Proprietary quantization technology that achieves an average of 3.6x inference speedup and 4x memory reduction. Their 'full-stack' understanding of how AI models interact with silicon enables optimization without significant accuracy loss.

How they differentiate

SqueezeBits specializes in ultra-low-bit quantization (down to 4-bit or less) without performance loss, leveraging hardware-aware quantization-aware training (QAT). Unlike Nota AI, which focuses on automated pruning platforms (NetsPresso), SqueezeBits emphasizes full-stack AI optimization for next-gen NPU ecosystems and Large Language Models (LLMs).

Main competitors

Nota AI, Neural Magic, Deci (Acquired by NVIDIA)

Key partnerships

Rebellions: Strategic partnership for NPU ecosystem expansion and full-stack AI optimization., Intel: Collaborated on optimizing LLMs and diffusion models for Gaudi-2 NPUs., Qualcomm: Integrated OwLite with Qualcomm AI Hub to support on-device AI applications., Strategic Investors: Kakao Ventures, Samsung Next, Naver D2SF, POSCO Capital.

Notable customers

Naver, SK Telecom, Qualcomm (Partner), Rebellions (Strategic Partner)

Major milestones

Secured Seed investment from Naver D2SF and POSTECH Holdings in May 2022, Launched OwLite, a SaaS-based AI quantization toolkit, in late 2023, Closed 2.5B KRW Pre-Series A round led by Kakao Ventures in Jan 2024, CEO Hyungjun Kim named to Forbes 30 Under 30 Asia 2024 list, Established strategic 'full-stack AI' partnership and investment with AI NPU leader Rebellions in Dec 2024

Growth metrics

Successfully completed technical verification (PoCs) with South Korean giants Naver and SK Telecom; recognized in Forbes 30 Under 30 Asia 2024.

Market positioning

Deep-tech AI infrastructure and model efficiency specialist, bridging the gap between massive AI models and resource-constrained edge/on-device hardware.

Geographic focus

South Korea, Asia-Pacific, North America

Patents and IP

Holds multiple proprietary patents in AI model quantization-aware training (QAT), post-training quantization (PTQ), and deep learning accelerator architectures.

About Hyungjun Kim

Hyungjun Kim is a specialist in AI model compression and quantization. He earned his PhD from POSTECH, focusing on deep learning acceleration. Under his leadership, SqueezeBits has secured investments from top-tier VCs like Kakao Ventures and Samsung Next. He was recognized in the Forbes 30 Under 30 Asia 2024 list for Deep/Enterprise Tech.

Official website: