Amol Deshpande’s Frontera Is Radically Innovative, Owning ABA Clinics and Its Proprietary AI

A $5,500 diagnosis that took eight weeks

DENVER, COLORADO. In 2014, a 2-year-old boy was nonverbal, walked on his toes, and did not respond when his parents called his name. His father, a general partner at Kleiner Perkins Caufield & Byers who was about to leave to co-found a new agricultural data company, paid $5,500 out of pocket for a private autism evaluation. The diagnostic report came back roughly eight weeks after the process started.

The father was Amol Deshpande. The new company was Farmers Business Network, which Deshpande co-founded with Charles Baron in 2014. At Kleiner Perkins he had risen from associate to general partner and led the firm’s first investment in Beyond Meat. Farmers Business Network (FBN) went on to raise more than $900 million in venture funding and reach a roughly $4 billion valuation by late 2021. Deshpande stepped aside as CEO ahead of starting Frontera.

What he was building toward, in retrospect, was a company that would not exist for another nine years. In 2021, Deshpande launched Divergent Investments, a family-office initiative that committed an initial $10 million across early-stage companies in neurodiversity, healthtech, and edtech, with portfolio names including the virtual-reality social-skills startup Floreo and the learning platform Learnfully. Two years later, in 2023, he founded Frontera Health.

In February 2025, Frontera came out of stealth with a $32 million seed round co-led by Lux Capital and Lightspeed Venture Partners, with Bison Ventures, Menlo Ventures, and Inspired Capital participating. The company is headquartered in Denver. Its co-founder is Jeff Flores. The AI and machine-learning function is led by Manu Kohli, Ph.D., a co-founder of the autism-care technology company CogniABle.

Owning both sides of the build

In ABA, the dominant business models have professionalized around two centers of gravity. Private equity has rolled up clinics into multi-state platforms (Hopebridge, BlueSprig, ABS Kids, Caravel, Behavioral Innovations, and Action Behavior Centers among them) and brought operational discipline to a fragmented field. Independent software vendors have built practice-management, scheduling, billing, and electronic-records tools that those clinics now run on. Each model has worked in its own way, and each has pulled significant capital toward the category.

Frontera is doing something different. Deshpande’s structural bet is that the next layer of value in ABA, the layer that turns clinical data into faster and more personalized care, cannot be built without operating the clinical environment that produces the data. So Frontera owns clinics. Five physical locations across New Mexico run under a sub-brand called FronteraCare, and the company says it now reaches families in more than ten markets statewide. They are real ABA practices with real waitlists and real BCBAs on staff. They are also the live R&D environment where Frontera’s AI gets trained, tested, and broken in before it ships to outside customers.

The clinics sit in Las Cruces, Santa Teresa, Hobbs, Clovis, and Santa Fe, communities that rarely show up on the expansion maps of the larger ABA platforms, and a recent push into northern New Mexico now extends into Española and Taos. FronteraCare offers in-clinic, in-home, and virtual ABA, plus online autism diagnosis, so some families are served without a brick-and-mortar location nearby. The clinics also partner with New Mexico schools to provide onsite support for teachers and families, a way to reach children who might never arrive through a traditional clinic referral. The Denver technology team builds against and learns from what those clinics see every day.

The supply problem Frontera is built around

The conditions Frontera is responding to are the same conditions every ABA operator is now facing, and sharper at the rural edge. The CDC’s most recent Autism and Developmental Disabilities Monitoring Network report, released in April 2025, put U.S. autism prevalence at 1 in 31 children, up from 1 in 36 in the prior reporting period. The BCBA workforce has grown but not nearly fast enough to match. Frontera’s own company materials cite that 12% of BCBAs work in rural areas and that 21% of children needing care sit on waitlists of a year or longer.

Deshpande has framed the math directly. “There are three times as many kids who need intervention relative to those who get it. Five times in rural markets,” he said in a February 2025 essay by Bilal Zuberi, the Lux Capital partner who co-led the seed investment. “We cannot ask clinicians simply to work more hours. We need innovation.”

Frontera’s strategic response has been to build where access is worst. Across its New Mexico communities, FronteraCare now has roughly 230 families in active care and is completing about 70 evaluations a month, a figure the company expects to push toward 100 as capacity grows. It says it has cut the wait between a completed evaluation and the start of care from about 50 days to 15.

The New Mexico clinics give the company something most pure-play ABA-tech vendors do not have: a continuous stream of real therapy sessions, assessments, diagnostic interviews, and BCBA workflows to develop and test against. The clinics also give Frontera a credibility argument when talking to outside ABA operators. The case the company makes is that the products work because the people building them are sitting next to the BCBAs who use them every day.

What the Frontera stack actually does

Frontera’s product line is organized around three products. The Assessment Builder is the most heavily marketed of the three. A full assessment can run eight to ten hours; the tool targets report writing specifically, which the company says falls from about four hours to roughly two, a 50% reduction in clinician time. The Diagnosis Builder targets a similar compression for the diagnostic report a clinician produces after evaluating a child for autism. In an early-2025 case study with its partner AT4K, Frontera reported the diagnostic write-up dropping from three hours to one. Reports generated through the two builders combined grew about 150% from January to May 2026, with May on track to exceed 1,300 completed reports.

The third product is the technical bet at the center of the company. Digital Phenotyping uses multimodal AI (video, audio, and text) to analyze a therapy session at 30 frames per second, tracking eye gaze, verbal and nonverbal behaviors, physical engagement, and other signals that have traditionally been captured through clinician notes and recall. The system produces structured behavioral data the BCBA can then incorporate into a treatment plan or use to refine an assessment.

Frontera is careful to position the underlying technology as something more than a thin layer over a general-purpose language model. The company’s materials emphasize that the Assessment and Diagnosis Builders run on what Frontera describes as thousands of pages of clinician-designed prompt engineering, developed in collaboration with practicing BCBAs and clinically validated in the company’s New Mexico clinics. In the company’s own phrasing, the product is a vertical AI system, not a ChatGPT wrapper.

“The goal of AI is not to replace clinicians, it’s to provide leverage to scarce clinical resources so more kids can get critical interventions.” Amol Deshpande, Founder and CEO, Frontera Health (2025)

What Frontera does not claim. The company has not positioned itself as the largest ABA-tech vendor, the lowest-priced option, or a replacement for clinical labor. The value proposition, repeated consistently across investor materials, podcast interviews, and the company’s own about page, is force multiplication: AI as added capacity for scarce clinical hours, not a substitute for them.

What comes after rural New Mexico

The first significant test of whether the Frontera model travels sits outside the company’s own clinic footprint. In January 2026, Frontera announced a strategic partnership with Catalight, a California-based nonprofit that operates one of the largest behavioral health provider networks in the country. Catalight’s network spans more than 16,000 practitioners serving 24,000 clients and families annually. Under the partnership, Catalight will deploy Frontera’s AI tools across its network to shorten assessment and documentation time, and to expand parent-mediated care delivery. Eight pilot organizations were live as of mid-2026, with a broader network rollout set to begin in late June 2026.

The deal matters because it puts Frontera’s products into a clinical environment Frontera does not own. The data flywheel that gives the company its differentiating asset, a continuous stream of real session video and clinical notes from the New Mexico clinics, does not extend automatically to the Catalight network. Inside that network, Frontera is a vendor, not an operator. Whether the products work as well in someone else’s clinical workflow is the empirical question the partnership will answer over the next 12 to 18 months.

Deshpande has been working on the neurodiversity thesis for longer than Frontera has existed. Divergent Investments, the fund he launched in 2021, has been a steady source of early-stage exposure to companies in the same orbit. Lightspeed’s investment memo on Frontera credits Deshpande with more than 30 investments since his son’s diagnosis, many of which the firm says shaped his understanding of where the field’s real gaps sit.

Eleven years after the eight-week wait for his son’s diagnosis, Deshpande has built the company he says did not exist when his family needed it. Whether the model holds outside the New Mexico clinics where it was developed, and how cleanly the AI performs inside the workflows of a 16,000-practitioner partner like Catalight, are the questions the next phase of Frontera is built to answer.

AT A GLANCE

SOURCES & REFERENCES