Founder JD Ambati is extending his robotics bet from picking recyclables to coordinating belts, sorters and maintenance, with operators approving every action.
By RuntimeWire Staff · Published
Primary source: PR Newswire
Why it matters #
Navigator moves EverestLabs beyond selling robotic sorters and toward owning the software layer that coordinates an entire recycling plant. That creates a larger market, but puts Ambati against industrial incumbents already connecting AI analysis to sorting equipment.
EverestLabs founder JD Ambati is moving his eight-year-old recycling automation bet beyond robot arms. In an August 24 announcement, EverestLabs launched Navigator, an AI system designed to watch material moving through processing plants, diagnose operational problems and prepare changes to equipment settings for an operator to approve.
Ambati started EverestLabs in 2018 after a family trip to India exposed him to sprawling landfills and waste systems that had seen little of the computer vision and automation arriving elsewhere in industry. His wife challenged him to apply his background to the problem, according to his account to TechCrunch. Ambati, who studied chemical engineering before earning a master's degree in computer science focused on AI, had previously worked across product, sales and strategy at Rocket Fuel, Fandom, Martini Media and 24/7 Real Media.
That combination shaped EverestLabs from the start. Ambati understood the software, but he also saw an industrial process built around conveyor belts, aging machinery and people sorting fast-moving waste in punishing conditions. In a 2021 interview with investor Sierra Ventures, he argued that falling computing costs had finally made robotics practical in complex, overlooked environments.
Navigator is the next expression of that thesis. EverestLabs spent its earlier years building RecycleOS and compact robotic sorting cells. Ambati, now founder and chief strategy officer, is trying to turn the data generated by those systems into a control layer for the rest of the plant.
From robotic hands to a plant-level brain
Navigator combines edge computer-vision models, vision-language models and specialized software agents. Cameras classify objects moving on belts. The vision-language layer interprets shifts in composition or contamination. Four agents then handle data queries, vision alerts, equipment settings and fleet health through a conversational interface.
An operator could ask where a plant is losing valuable PET plastic, for example, and receive a diagnosis tied to a particular line, shift or piece of equipment. Navigator can then prepare a recovery plan or equipment preset.
EverestLabs is explicit in its product documentation that Navigator "never acts autonomously." Equipment adjustments require operator approval, including changes passed to industrial control systems through SCADA. That safeguard narrows some of the autonomy language in the launch release and a Caglia Environmental customer testimonial describing a "fully AI run and managed plant."
The human approval step also makes Navigator easier to introduce into facilities where an incorrect belt-speed or sorter-setting change could halt production, damage equipment or degrade the value of a finished material bale. EverestLabs says Navigator connects to legacy controls without forcing operators to replace existing lines or stop production for extensive retrofits.
Caglia Environmental is an early customer. EverestLabs also named Schneider Electric and sorting-equipment maker Pellenc ST as integration partners. Those relationships matter because Navigator needs access to plant controls and machinery if it is to progress beyond another analytics dashboard.
The financial claims need operating evidence
EverestLabs says Navigator can raise throughput by reducing uneven material flows and idle belt time. The launch release estimates a 20% to 30% increase in facility capacity. The Navigator product page makes a broader claim of 20% to 40% higher throughput and says a 400-ton-per-day facility could generate $2 million to $4 million in additional annual revenue.
Those figures are EverestLabs estimates rather than independently audited results. The difference between the release and product-page ranges also shows how much depends on a facility's baseline utilization, equipment and incoming material. A plant already running close to capacity has less room for an AI system to find.
EverestLabs makes similarly large claims around predictive maintenance. The Navigator product page estimates roughly $1 million in annual savings from preventing downtime, while EverestLabs says typical facilities lose $100,000 to $800,000 in recoverable material to landfill each year. Navigator's ability to identify where that material escaped is technically plausible; the size of the financial return will depend on commodity prices, plant configuration and whether operators implement its recommendations.
Ambati is selling against a concrete expense rather than an abstract AI budget. Every valuable can, plastic container or piece of metal sent to landfill represents foregone revenue, followed by the cost of disposing of it. That gives EverestLabs a straightforward path to proving value if Navigator's measurements hold up in production.
The market already has an agent race
EverestLabs calls Navigator the first multi-agent AI platform built for materials processing and recycling facilities. The broader first-mover claim needs narrowing.
In May, TOMRA introduced an AI-native platform from PolyPerception that lets recycling operators query plant data in natural language, generate reports and create operational alerts. TOMRA also increased its PolyPerception stake to 51%, pairing the software with an incumbent supplier of industrial sorting equipment.
Navigator's sharper differentiation is EverestLabs' attempt to combine material vision, robotic fleet monitoring, plant analytics and operator-approved control changes in one interface. EverestLabs is entering that contest with its own installed robotics base and data gathered from physical sorting lines. EverestLabs says its models have been trained on more than 100 billion objects, while its systems have transformed more than 200 lines and logged over 10 million belt hours. Those deployment figures are self-reported and undated; EverestLabs also reports them on its homepage.
Ambati's second act at EverestLabs
Ambati handed the CEO role to R. Paul Singh in 2026, according to Singh's Forbes profile. Singh had been an early EverestLabs investor and adviser before taking the job, leaving Ambati focused on product and strategy. The arrangement gives EverestLabs a founder working on the technical and industrial thesis while an operator with several prior exits manages the next stage of commercialization.
EverestLabs last disclosed a major financing in 2022, when TechCrunch reported a $16.1 million Series A led by TransLink Capital. NEC Orchestrating Future Fund, Benhamou Global Ventures, Sierra Ventures, Morado Ventures and Xplorer Capital participated, bringing verified funding at the time to $24.63 million.
Navigator gives that capital-intensive robotics business a software layer that can reach beyond each individual picking cell. The sales case rests on whether EverestLabs can convert camera data into reliable plant decisions without asking operators to surrender control or rebuild their facilities.
Ambati described Navigator as "an AI process engineer, a data analyst and a controls specialist" for plant operators. That is an ambitious job description. The early deployments now have to show that the agents can improve an industrial process measured in tons, downtime and recovered commodity value, rather than the number of questions answered in a chat window.