Founder Spotlight articles are editorial profiles drafted with AI assistance and reviewed by Marc Hoag. They are based on publicly available information about the featured company and its founders. Spotted an error or want a correction? Email me.
Startup: TheCompressionCompany.com
Founder(s): Michael Stanway, Joe Griffith
Stage: Pre-Seed
Location: San Francisco
Enterprise sensor networks generate more data than they can move. A single satellite pass produces gigabytes. A fleet of autonomous vehicles streams terabytes per day. Medical imaging systems, industrial robotics, drone networks, all drowning in their own output. Traditional compression algorithms were built for static files. They can’t learn what matters in a stream of multispectral video or LiDAR point clouds. The result: massive infrastructure costs, latency bottlenecks, and analytics pipelines that operate on stale or downsampled data because the raw feed is simply too large to process in real time.
The problem they’re solving
The Compression Company addresses a specific infrastructure problem that grows worse as sensor hardware improves. Better cameras and faster sampling rates don’t help if the data can’t leave the device or cross the network in time to be useful. Enterprises are caught between throwing away data (and losing signal) or building expensive pipelines to handle it. Standard compression — whether JPEG, H.264, or lossless algorithms — treats all pixels equally. They can’t distinguish signal from noise in domain-specific sensor streams. A neural codec trained on satellite imagery or medical scans, by contrast, learns what downstream tasks actually need. It discards noise intelligently and preserves what matters, achieving 50x compression in some benchmarks while keeping reconstructed data visually indistinguishable from the original.
The company’s approach runs at the edge, on the device itself, before data ever leaves the sensor. No hardware changes required. Just a software layer that learns from your specific data and adapts as sensor conditions shift.
Why these founders
Michael Stanway and Joe Griffith come from deep backgrounds in signal processing and systems engineering, the kind of foundation that makes neural compression possible without magical thinking. Stanway has spent years working in data infrastructure and sensor systems; Griffith brings expertise in building production AI systems that run on constrained hardware. Neither is new to the problem space. They’ve both worked on the operational side of large-scale data pipelines, which meant they saw the compression problem not as an academic exercise but as a recurring cost center that killed margins. That operational context shows in the product.
Why we’re watching
The legal surface area expands quietly as sensor networks grow. From satellite imagery to drone footage and medical scans, compression directly affects privacy and compliance. Compress poorly, and you’re transmitting unnecessary detail that should never leave the device. Compress carelessly, and you risk downstream liability: if a reconstructed image contains artifacts that bias an AI diagnostic or autonomous decision, who bears responsibility? There’s also IP strategy to consider. Custom-trained models become proprietary assets, but they’re also tied to specific hardware and data regimes. As adoption spreads, data processing agreements (DPAs) with cloud providers, edge infrastructure vendors, and customers all hinge on where compression happens and what gets retained.
The company’s February funding round — $3.4M lead by Long Journey, an early investor in SpaceX and Uber — signals confidence in the model. As they scale into regulated verticals like medical imaging, autonomous vehicles, financial surveillance, the governance layer around AI-native compression becomes material.
Hoag Law.ai provides flat-rate fractional general counsel services for AI-native startups, covering SaaS contracts, privacy compliance, AI governance, and startup IP strategy. Learn more →
