California Waste Solutions upgrades AI scanning at Bay Area MRFs
California Waste Solutions is deploying two technology upgrades at its Bay Area material recovery facilities: expanded Pellenc optical-sorting capacity at the Timothy MRF in San Jose, and new AI-powered scanning units from EverestLabs at the 10th Street MRF in Oakland. The projects target different points of the same operating challenge, namely improving the rate of material recovery and the purity of sorted commodity streams.
The Oakland installation is the more novel of the two. EverestLabs scanners will observe material entering the container line and generate a continuous item-level record of what reaches the end of the process headed for disposal. The system will also report on belt burden depth and exposed belt surface, giving operators data on how material density varies across the line. California Waste Solutions says the primary purpose is to give sorting teams a richer picture of operating conditions, not to remove human judgement from the floor.
The operational logic
At the Timothy facility in San Jose, the company is increasing the number and capacity of existing Pellenc optical sorters. Optical sorting relies on near-infrared and other sensor technologies to identify and deflect target materials at high throughput speed. Expanding that capacity improves both the volume of recoverable material captured and the cleanliness of the finished commodity bale, two performance metrics that do not always move together.
In Oakland, the EverestLabs deployment is explicitly positioned as a data foundation for future capital decisions. California Waste Solutions said it is evaluating robotic sorting, additional optical sorting, or a combination of the two as a possible next phase, and intends to use scanner data to identify where an additional recovery technology would produce the greatest operating benefit. Michael Duong, president of the company, said the approach was designed to keep the firm "ahead" in sorting technology as AI capabilities mature.
Market context
The application of machine-vision and AI to materials recovery is a growing sub-sector within industrial automation. EverestLabs is among a cohort of startups applying continuous computer-vision inference to recycling infrastructure; others active in the space include AMP Robotics and Greyparrot, which similarly generate item-level data from conveyor lines to inform sorting decisions. The competitive premise across this category is that legacy MRF measurement, which relies on production totals and periodic quality checks, leaves recoverable material losses invisible until they accumulate at scale.
Optical sorting has a longer commercial history. Pellenc ST, the French sensor-sorting manufacturer whose equipment California Waste Solutions has operated since the mid-2000s, competes with Tomra and Steinert in the near-infrared and multi-sensor sorting segment. Capacity additions at existing installations are typically lower-risk investments than new-technology pilots, which is consistent with California Waste Solutions treating the two sites as complementary rather than interchangeable upgrades.
Regulatory and commercial backdrop
California's recycling infrastructure operates under some of the most stringent material-quality mandates in the United States. SB 1383, the state's short-lived climate pollutants law, imposes organic waste diversion targets on jurisdictions, while CalRecycle sets contamination thresholds that directly affect the commercial value of commodity bales. Improving purity through better sorting capacity and tighter line monitoring has a direct bearing on whether sorted material meets those thresholds and can be sold into commodity markets rather than landfilled.
California Waste Solutions, founded in 1992 and still family-owned, holds municipal collection and processing contracts in Oakland and San Jose. The company did not disclose the capital cost of either upgrade or a timeline for completing the installations. The Oakland scanner project's next phase, evaluating robotics, will be an indicator of whether AI-generated conveyor data translates into measurable recovery rate improvements in commercial MRF conditions.