TheraDriver founder Malavica Sridhar left Waymo to solve ABA therapy’s turnover problem and found the answer buried in one-star reviews. Her scheduling-first platform now works with hundreds of providers and is expanding into speech and occupational therapy.
Key Takeaways
- ABA’s retention crisis is really a scheduling problem. A mass scrape of provider reviews showed that scheduling, not pay, was the top driver of behavior technician complaints, and TheraDriver built its first product around that single insight.
- The wedge was a narrow, daily pain point. The company led with morning cancellation coverage so no client gets canceled on, and that feature opened the door to a broader workforce analytics platform.
- The AI treatment-planning tool assists the BCBA without replacing clinical judgment. It drafts administrative content automatically and offers clinical suggestions with confidence scores, but clinicians approve every recommendation and set therapy hours themselves.
- Expansion into adjacent therapy markets is deliberate, not opportunistic. Sridhar is extending into ABA-adjacent speech and occupational therapy first, with home health as a longer-term target, while keeping the team focused on a small product set.
Malavica Sridhar did not come up through behavioral health. McKinsey first, then San Francisco in 2017, then a startup building machine learning for private-market investors, and finally the machine learning infrastructure team at Waymo, Google’s self-driving car unit. She’s always wanted to build something from the ground up and building in healthcare gave her the opportunity to build with a strong mission in a space craving technology.
Autism care got on her radar through her family and her friends. A cousin has autism and ADHD; meanwhile, several people she knew were opening ABA clinics in different parts of the country. The more she looked at the technology these clinics were buying, the more one fact jumped out: the market was drowning in electronic health records. “I went to a handful of conferences when we first started,” she said, “and it’ll have 10 booths of EHRs that are all claiming to be all in one.” Competing in that crowd held no appeal. But underneath the EHR noise was a complaint she heard again and again, about turnover, and that was the thread she decided to pull.
Her method was the one a machine learning engineer would reach for. She scraped the public record on as many providers as she could find and ran the one- and two-star Glassdoor and Indeed reviews through an AI summarizer to see what technicians were actually angry about. The ranking surprised her. Scheduling came first. Training second. Pay third, culture fourth. That order is easy to get wrong, she points out, because an operator taking a single exit interview will hear a technician gripe about money and file it under pay, when the real problem is that the schedule kept shifting. The pattern tracks with what other reporting has found, that the ABA workforce shortage is driven more by retention than recruitment. As one version of the complaint put it: “I wanted a certain schedule and I didn’t get it, so I need to go somewhere that gives me a consistent schedule.”
Before she hired anyone, Sridhar mocked up the product in Figma and pitched it around as if it already existed. The fake software worked. Once prospects began asking about contracts and what it would cost, she had the only validation that matters: people were willing to pay. From there it moved quickly, a team, a venture round, and, she says, a first product live in January 2024. Today TheraDriver counts hundreds of providers as customers, among them names the industry will recognize: ACES, Proud Moments, Helping Hands Family, Stride Autism Centers, InBloom, and First Steps. ACES and Proud Moments rank among the larger national ABA providers reshaping autism care. Almost all of that growth came through word of mouth, which is how things tend to travel in a referral business.
The Wedge: Solving the 6 a.m. Cancellation
TheraDriver’s first feature did exactly one thing. Behavior technicians work long sessions, day after day, and a workforce that size cancels constantly. Sridhar’s analogy is an HVAC company that loses a plumber at the last minute with a customer already waiting at home. The difference is that the customer here is a child, and the failure mode is a kid losing a therapy session.
So the company built coverage reshuffling for early-morning cancellations. A technician calls out, the software finds a replacement, and the client never feels the gap. Internally they called it the swap function for 6 a.m. cancellations. What it replaced was a color-coded Excel sheet and a scheduler trying to fit the pieces together before the doors opened at eight, the kind of manual scheduling, billing, and intake work that still runs much of the ABA back office.
Then the product grew along the contours of the problem. Next came prescription fulfillment. A BCBA authorizes a client for a certain number of hours over an authorization window, typically six months, and that figure is a clinical decision, not an arbitrary one. In practice almost every provider falls short of it, worn down by thin scheduling, cancellations, and churn. TheraDriver started showing clinics where each client stood against the authorized total and nudging them to act, schedule more, or call the parent, before the gap grew too large to close.
Utilization was the third piece. A surprising number of clinics, small clinician-owned shops and large groups alike, cannot say what their technician utilization actually is, let alone how to lift it, and the question gets sharper when an owner has to decide between adding hours for current clients and signing new ones. Answering it turned into a full operations analytics dashboard of the kind many ABA and SUD operators still struggle to build: prescription fulfillment, technician utilization, net cancellation rate, and the distance between a technician’s scheduled hours and the hours they say they want. That last number loops back to where the company started. Technicians who do not get the hours they want are the ones who leave.
The ROI Math, and What the Sales Pitch Leaves Out
With operators, Sridhar argues from the spreadsheet. The pitch hinges on lifting client prescription fulfillment, which in turn rests on two levers, technician utilization and net cancellation rate. TheraDriver keeps an ROI calculator that, by the company’s own figures, pencils out to a return of 12 to 15 times the software’s cost against a provider’s top line. Sometimes she just hands prospects the calculator and lets them enter their own numbers and their own contract price.
The interesting part is what the sales pitch downplays. Retention, the thing that launched the company, is a lagging indicator. Six or seven months can pass before a frustrated technician actually quits, and a slow-bleeding loss is harder to sell against than one a provider feels every morning. “A lot of buyers don’t buy the product because of behavior technician retention,” Sridhar said. The pain arrives too late. So TheraDriver sells on the daily math instead, fulfillment and utilization, and lets better retention show up later as a byproduct of finally giving technicians a schedule they can count on.
AI Treatment Plans: Automation Versus Recommendation
The treatment-planning tool is where TheraDriver has to answer for the thing clinicians are wary of, which is AI getting anywhere near clinical work in ABA. Sridhar splits the tool in two: automation and recommendation.
Automation covers the dull material, a client’s birthday, the physician’s name, the parent’s name, all assembled into a first draft. “The birthday is the birthday,” she said. “There’s no prediction.” Recommendation is handled with more caution. For behavior reduction and skill acquisition, the two sections BCBAs guard most closely, the tool offers a bank of suggestions, each tagged with a similarity or confidence score drawn from comparable client profiles, and the clinician has to accept or reject them one at a time. A BCBA might see an item flagged at 99 percent similarity to children with comparable assessment scores, nod, and accept it, or look at a lower score and pass.
The model comes straight from consumer AI. Tools like Claude, Gemini, and ChatGPT caught on, Sridhar argues, precisely because they hand you a draft rather than a verdict; you tweak the email before you send it. TheraDriver’s treatment plan works the same way, editable by prompt, with an approve button that quietly feeds a reinforcement-learning loop. And on the single most loaded number, weekly direct therapy hours, the tool stays silent. “You need to put that stuff in,” she said. The clinician decides.
Parent Engagement, Partnerships, and a Disciplined Expansion Path
Parent engagement, the fourth product, is the one Sridhar will openly admit has given her trouble. She is, in her words, obsessed with it, and generous about a competitor, crediting Frontera Health, the AI startup Catalight tapped to scale its lower-intensity care model, with actually cracking parent adoption. TheraDriver threw a lot at the problem and watched most of it fail. Virtual parent coaches. A BCBA-on-demand setup promising a text reply inside 15 minutes, borrowed from what Summer Health does with pediatricians. None of it held, because parents would not trade their time. Ten minutes, 15, half an hour for training: the answer was no.
What worked was asking nothing of them. Through a partnership with Rethink, TheraDriver now builds an AI session summary, plain English, no jargon about differential reinforcement, that tells a parent the five things their child worked on that week and five ways to reinforce them at home. The reference point Sridhar reaches for is Brightwheel, the daycare app that pings parents a snapshot of their toddler’s day. Clinically, the payoff is generalization, skills carrying over from the center to the living room, and that happens to be a selling point for providers too.
She is in no rush to expand. The scheduling engine, she explains, only works where three conditions hold: high-frequency care, a workforce that cancels a lot, and rules about which staff can be paired with which clients. Plenty of industries fit. Home health is the big one. Rather than leap, though, TheraDriver is widening into the disciplines next door first, speech and occupational therapy, and treating anything further out as a someday. The code might port over; the sales motion and distribution will not, and scattering the team’s attention across new markets is the mistake she is most determined not to make.
