Artificial Intelligence

Whistleblowers charge CEO of NJ firm with inflating AI capability, calling employees “dirty Indians”

They allege that ElectrifAi is touting an improperly-developed healthcare AI tool amidst a pandemic and accuses the CEO of racism, a charge that finds corroboration in an employment lawsuit. The company pushes back saying the product is being used by unidentified healthcare entities, is listed on the Microsoft Azure marketplace, and allegations against CEO are simply untrue.

Edward Scott took over as CEO of ElectrifAi in late 2018.

For years, Opera Solutions did business as a data analytics consulting company helping clients both in and out of healthcare. They include the Center for Medicare and Medicaid Services, where the company’s data analytics chops helped them score an operational analytics and fraud surveillance contract.

But financial mismanagement led creditors to come calling and ultimately one of them — a private equity firm based in San Francisco called White Oak Global Advisors — took over the company. In late 2018, Edward Scott, managing director at White Oak, was named CEO of Opera. [His LinkedIn profile has been amended recently to reflect that his tenure at White Oak ended in September 2018.]

From then on, the company has been on a relentless makeover journey to present itself as an artificial intelligence, products company. It even got a new name — ElectrifAi (pronounced electrify). When Covid-19 emerged into the U.S. consciousness, the company began touting a product in the hospital world. But people who call themselves “whistleblowers” and decline to publicly identify their relationship with ElectrifAi say the emperor has no clothes. They allege that ElectrifAi (or Opera) has zero clinical experience historically and has never worked with imaging data, which is the very basis for the company’s PulmoAi product. But over the last six months, Scott and ElectrifAi’s senior executives have touted PulmoAi, which has no FDA clearance or approval, saying it can quickly diagnose Covid-19 in emergency rooms and help doctors make decisions about who needs hospitalization and who can be safely sent home.

The whistleblowers believe that PulmoAi and the way that it is being marketed is dangerous to people during a pandemic. Aside from their troubling allegations about the product, the group also claims that Scott’s managerial style has led to clients to stop renewing contracts.  Further, Scott’s racist attitude toward employees of Indian descent and other minorities has left employee morale at a nadir. An employment discrimination lawsuit filed on Aug. 24 corroborates the whistleblowers’ characterization alleging that Scott repeatedly called Indians, “dirty” and those in leadership positions that he fired after taking over as “the Indian Mafia.” 

For their part, the whistleblowers hope for Scott to be replaced by someone more capable, respectful and honest. Through a crisis communications expert with a formidable reputation, the company declared that PulmoAi is legitimate, is being used by unnamed healthcare entities, and is featured in Microsoft Azure’s cloud marketplace. ElectrifAi also denied all charges of racism, discrimination and misogyny as categorically false. 

“Automating early diagnoses of Covid-19”
In March, a public relations professional sent an attractive pitch — “Using computer vision, ElectrifAi has created a program that can detect pneumonia-like symptoms through CT scans of the lungs and are now working with hospitals worldwide to collect relevant images to further train the AI and fully automate early diagnoses of Covid-19 which presents similarly.”

Wow.

Rapid testing wasn’t as yet easily available, so this would be a phenomenal tool to have in ERs. But AI needs a lot of training and tons of data points. Given that Covid-19 was still relatively new, how could they get enough CT-scans to be able to train the data and distinguish it from pneumonia?  The PR person responded by borrowing from a post on the company’s LinkedIn page.

Using ElectrifAi’s “minimal model” technology, they’re able to train an AI with no more than 100 2D annotated images of various cross-sections of the lungs.

Wow again.

The public relations professional was pitching both an interview with Scott and a neuroradiological expert at Catholic Health Services of Long Island, a hospital group in New York. At the time, Catholic Health Services was just about to implement ElectrifAi’s PulmoAi technology meant for X-Rays in emergency rooms — any day now, I was told.

So a few weeks later, MedCity News published this story in early April that attempted to probe Scott’s declarations. In the interview, he provided no details about the technology, named no European hospitals or other overseas hospitals where he said the AI was trained and/or deployed. And Dr. Craig Sherman, service line director of Neuroradiology at Catholic Health Services, while “impressed” by the company’s tool and eager to deploy it imminently,  acknowledged he was a paid consultant of ElectrifAi. As soon as the bureaucratic paperwork could be dealt with, Sherman would be sending over X-Rays to ElectrifAi.

Five months after that interview with Sherman and Scott, it’s not clear whether Catholic Health Services of Long Island actually ever used ElectrifAi’s PulmoAi. After inviting me to connect on LinkedIn once the initial story was published in April, Dr. Sherman recently did not respond to messages. He did not respond to a call either.

Separately, after several weeks of unanswered voicemails and messages to media contacts at the health system, one responded:

Thanks for your inquiry. Please connect with those who you spoke with for your original story. – Greg Sleter, Public Relations & Brand Reputation Manager, Catholic Health Services

Sleter didn’t respond to a follow-up email.

“We are unapologetic for our innovation”
After the MedCity story published on April 6, a somewhat defensive and lengthy blog post emerged from ElectrifAi a few weeks later. The company’s main argument that they return to throughout the piece is that it’s just pure math, 1+ 1=2 and they are simply trying to help people.

John Quackenbush, Henry Pickering Walcott professor of computational biology and bioinformatics and chair, Department of Biostatistics, at the Harvard T.H. Chan School of Public Health, decoded the post. He found the arguments,  “pretty nonsensical.” In an email, he wrote:

  • They claim they trained their algorithm with 10 Covid patients. Anyone doing machine learning will tell you that [the] sample size is far too small to draw any meaningful conclusions. They might argue that they distort the images and that they add noise, so the effective number is larger, but the question is whether this captures the diversity of images that one would see in practice. The answer is simply that ten of anything this complex isn’t enough. Will it work on a Siemens or a Phillips or a GE machine? At what settings? If the machine is new or old? If the patient has been on a ventilator? If they were a smoker? If they had emphysema?
  • [They] claim they only trained with Covid images. So the question is what did they train against? Simply put, if the algorithm is designed to answer the question, “Yes or no, does my patient have Covid?”, then you need examples of both “Yes” and “No.” If the algorithm is trained on a data set where the only answer is “Yes,” then the only answer it can give is “Yes.”
  • They state, “We have no interest in publishing research in an effort for self-promotion and peer accolades. Our sole interest to open up the technologies so they can help people around the world.” But they never make their code available and they haven’t published anything, so nobody has the opportunity to actually review what they have done and they certainly haven’t “opened up” any technology.

While ElectrifAi is not interested in publishing research for that would be self-promotional, it appears that executives don’t have any qualms promoting PulmoAi in non-academic, non-peer-reviewed settings. Between March and May, Scott and other executives have appeared in at least five podcasts including the Empowered Patient podcast, THINK Business with Jon Dwoskin, one hosted by AI recruiting and staffing company Alldus, and a YouTube video. Over a five-month period as of mid-August, the company had 14 LinkedIn posts specifically promoting its AI capability in healthcare, though healthcare is not the only industry it serves. There have been positive external news coverage as well. 

This marketing narrative including laudatory news articles is what the ElectrifAi whistleblowers want to counter. They believe that Scott sees the pandemic as an opportunity to be milked.

“The company has rapidly lost revenues since Ed Scott took over. There are hardly any new clients. Most existing clients have not renewed their projects with the company due to extremely combative new management,” alleged one whistleblower in an email. “[Pulmo AI] is one of those schemes to quickly showcase something seemingly disruptive and make a quick buck. This will not stand any reasonable questioning or unbiased peer reviews done by qualified personnel in the field.” 

Another added that historically the company has had no clinical expertise, let alone medical imaging expertise, that Scott et al are showcasing:

“ElectrifAi has had experience with text data — classification & clustering algorithms. We have never worked on any image related machine learning algorithms. Georgios Ouzounis was hired as an expert in the field but it seemed he wanted to use the company resources to create a product based on his Ph.D. thesis. There was no team in place and no tangible expertise in the analytics team for any sort of image recognition or image machine learning capabilities. Hence, any product using CT and X-Rays is just a marketing gimmick.” [Ouzounis is vice president of data science who has appeared in at least one podcast about ElectrifAi’s AI prowess in healthcare and Covid-19.] 

Another explained that in its previous incarnation — as Opera Solutions — the company developed a healthcare product, but not on the clinical side.

“Opera Solutions has a product for identifying missing charges in hospital bills. This was a result of an analytics consulting project that was productized. People who were involved in the development of the product have since left or have been fired.” 

This product is still in commercial use. Indeed the company’s website explains that it can help providers detect missed charges and boost revenue. In the March phone interview with Scott, I asked him to explain how a company goes from having billing, fraud detection and procurement software to diagnosing Covid-19 using CTs and X-Rays. It was a leap, I felt, and he bristled.

“We continue to innovate, Arundhati … and we are unapologetic for our innovation,” he declared.

It’s important to note that while the whistleblowers call into question the healthcare AI product, ElectrifAi has been accepted into the Inception accelerator program of a Silicon Valley AI heavy-hitter – Santa Clara-based Nvidia.

“Inception is a virtual accelerator program that supports over 6,000 startups using GPUs for AI and data science applications during different stages of product development, prototyping and deployment, but being part of the program does not mean the company’s technology or business model has been fully vetted by us,” said Hector Marinez, director of corporate communications at the firm, in an email.

“ElectrifAi is proud of the contributions it has made in the fight against Covid-19”
After several weeks of radio silence in which Scott did not return emails either sent directly or through the original public relations professional, response came from Michael Sitrick, the crisis management expert hired to ensure that ElectrifAi’s side of the story gets fair play. [Personnel at White Oak Global Advisors, which owns the company, and White Oak’s lawyers also didn’t respond to emails.]

Sitrick displays some of the not-so-complimentary media profiles of himself as the “Ninja Master of the Dark Art of Spin” prominently on the website of the company he founded and which bears his name. The Columbia Journalism Review describes his prowess and manipulative capabilities writing that “he specializes in crisis PR and has made a name for himself as the man you call when you have money and step in some s***.”[CJR spells out the word and this portrayal is also on Sitrick and Company’s website.]

Sitrick, only too aware of his reputation and how reporters are hard-wired to represent both sides, rushes in to say via email that he has blue-chip life science-industry clients like Amgen, Glaxo-Smith Kline, Anthem and Tenet Healthcare. But he and his firm have had less savory clientele too: Los Angeles Catholic archdiocese during the sexual abuse cover-up scandal; Harvey Weinstein until payments weren’t made, and R. Kelly.

No ElectrifAi executive was made available for an interview, but in an email from Sitrick, Jim McGowan, head of product at ElecrifAi, contended that PulmoAi technology has been “favorably received by healthcare providers inside and outside the U.S.” “A premier Cancer Institute” was “impressed;” ElectrifAi had “multiple calls with radiologists in [Italy], who showed intense interest.” But the company declined to name any institution or healthcare professional that it was working with.

McGowan said that ElectrifAi realized in mid-March that it’s “minimal model algorithm” doesn’t work with X-Rays but it was able to resolve the problem and this was tested by Microsoft and a “practicing radiologist” after which the product was successfully added to the Microsoft Azure marketplace.

“The product was granted two awards by Microsoft, including a $100,000 grant …,” said the email.

Only it wasn’t really a cash grant as the above suggests. An email from Shannon Fitzpatrick, senior healthcare partner development manager at Microsoft, that Sitrick forwarded when asked for proof, shows that it’s a credit. The email dated April 22 says Microsoft was offering ” a one-time Microsoft Azure monetary commitment credit to cover unexpected increases in Azure consumption that stem from the increased utilization of your solutions in the healthcare and life sciences.”

In other words, should people download PulmoAi from the Azure marketplace and use it so much that ElectrifAi’s cloud subscription fees increase, then Microsoft is willing to waive them for a year up to $100,000.

When reached by phone Fitzpatrick didn’t want to be quoted without express approval from Microsoft’s communications team. David Houlding, director of healthcare experiences who is responsible for healthcare products on Microsoft’s cloud platform similarly declined to comment. Neither Microsoft nor ElectrifAi would comment on how many PulmoAi downloads have occurred. But one thing is clear: Like Nvidia, Microsoft has not vetted the company’s Covid-19 AI technology.

McGowan also claimed that the company has communicated with the FDA to get an emergency use authorization but offered no proof of such communications — in fact, the blog post from late April shows that at least at that time, ElectrifAi officials never had any intention to go to the FDA.

These technologies are not “US FDA Approved.”

They could never be in the timeframe available to ever have an impact.

“Inadvertent exception”
If the technology is as valuable during Covid-19 as ElectrifAi executives claim, why is it that they won’t name any active users of the product or the radiologists and cancer centers it is working with? Sitrick has an easy answer: the company doesn’t have permission from those very individuals and institutions.

During the pandemic and attendant public health crisis, the healthcare industry has been more open than usual in naming technologies and platforms being used in the spirit of collaboration. Press releases are replete with tech vendors’ touting their tech being piloted here or deployed there, naming specific hospitals. ElectrifAi’s stance of complete secrecy is counter to that observed change.

The company’s reticence also flies in the face of the joint interview Scott and Dr. Craig Sherman, service line director of Neuroradiology at Catholic Health Services of Long Island did with MedCity News in the spring even before the X-Ray based PulmoAi tool was supposed to be pushed out to ERs across Catholic Health Services hospitals. [Recall that Sherman described himself as a paid consultant of ElectrifAi in the interview.] 

When asked to comment on this inconsistency Sitrick described is as an “inadvertent exception to the rule” of not naming clients without permission. [Which raises the question, did Sherman have the blessing of Catholic Health Services when he spoke to MedCity or was he speaking as a paid ElectrifAi consultant?]

“Opera is safe for White people Again”
Other than trying to peddle, what they believe is, a questionable AI product in a pandemic, the whistleblowers also charge that Scott engaged in threatening, racist, misogynistic behavior and introduced employee-tracking and monitoring products into the workspace. Their allegations are corroborated in an employment discrimination lawsuit filed on Aug. 24 by Aparna Kumar, a senior vice president at ElectrifAi. Kumar names not only various entities including ElectrifAi and White Oak Global Advisors as defendants but also Scott and Nancy Hornberger, executive vice president of revenue and co-head of healthcare in her civil complaint.

In the lawsuit Kumar, an Indian woman who became an American citizen in 2002, states that she was hired in 2008 and until recently was an account manager for ElectrifAi’s largest client, the Centers for Medicare & Medicaid Services, overseeing 34 employees. She describes her upward trajectory in the firm from business school intern to SVP, and then a sharp reversal of fortune when Scott took over as CEO in the Fall of 2018. She was demoted and her role taken over by white people, she charged, who were, in one case, recently hired. The complaint states that Scott used the words “dirty Indians” on multiple occasions as well as the term “Indian mafia” to refer to the company’s previous senior leadership. Here’s a specific example:

… in April 2019, Kumar and Scott had a one-on-one conversation in which Scott described a story of having rehired a former Company employee, who is White. Scott said to Kumar that he told the rehired employee, “I got rid of all the dirty Indians and it is safe for White people at Opera again.”

 

 

 

 

Scott fired hundreds of people of color, including Indians, and the racial animus appears to have been directed at employees of non-Indian descent as well. The complaint states that sometime in the spring of 2019, Scott fired a black woman who led an ElectrifAi team that contracted with Johnson & Johnson. Scott had previously referred to this woman as a “fat lazy Black bitch.”

If true, the allegation stands in stark contrast to ElectrifAi’s public posturing, where it has displayed support for the Black Lives Matter movement and separately for LGBTQ people on its LinkedIn page.

The complaint also notes how Scott had web cameras installed in the office seating area in the presence of employees, and in July 2019 instructed employees to install an application named Carbonite. This allowed him to replicate an employee’s computer hard drive and access websites that the individual visited. The company’s HR for a time was in essence just Scott – the complaint notes how the CEO “forced out” the head of human resources and the two remaining HR personnel resigned thereafter.

The lawsuit goes on to depict how over time, Kumar saw her role diminished and she was excluded from meetings with the very client — CMS — with whom she led the relationship on behalf of ElectrifAi. When folks from the CMS lauded her skill and expertise, Scott and others felt compelled to include Kumar in meetings only to sideline her thereafter. She cites numerous examples of Scott using the f-word in meetings and humiliating her, according to the complaint, while treating men very differently.  She was stripped of responsibilities last year and ultimately earlier this summer, Kumar was removed as the account manager for the HHS-CMS account and made Daily Project Director instead. She had held the account manager for the HHS-CMS account since 2016. While formally still a senior vice president, Kumar is supervised by Margaret Cox, a White woman who holds the title of vice president and now manages the HHS-CMS relationship, she asserts in the complaint.   

Kumar attempted to bring the discrimination and misogyny to the attention of White Oak Global Advisors and met with a lawyer hired by the private equity firm to air her concerns about her employment in 2019. That was Kathleen Einhorn, a lawyer from Genova Burns hired to investigate Scott on behalf of White Oak. When the hours-long interaction led to no discernible change – in fact she was removed from the Management Committee — Kumar then filed a complaint with the Equal Employment Opportunity Commission on or about Jan. 15, according to the complaint. The lawsuit was filed on Aug. 24. 

Einhorn did not return an email seeking comment. Neither did Pamela Hart who leads human resources at White Oak Global Advisors. Meanwhile, Kumar remains an ElectrifAi employee and is seeking punitive damages and other relief from the courts. Her attorney after initially responding to a call requesting an interview with Kumar stopped responding to subsequent queries.

Sitrick categorically denied Kumar’s allegations about Scott and his behavior, calling them “incendiary claims.” He provided what appear to be grainy screenshots of Kumar’s texts with Scott to disprove her depiction as a hapless employee victimized by a misogynistic, racist boss. This is from May 22, 2019, seven months after Scott joined as CEO:

The last line is One team, one Opera(the previous name of ElectrifAi.)

Another from Aug. 17, 2019

That was the last screenshot Sitrick shared showing Kumar being appreciative of Scott. Sitrick said that Kumar “remains one of the 5 highest paid employees in the company,” “about 40% of ElectrifAi’s workforce are women and that  “women hold 5 of the top 11 positions in the company.”

As mentioned above per the lawsuit, on “October 16, 2019, Kumar and three members of her team visited White Oak’s offices in Manhattan. They hoped to speak in-person to a high-level executive about Scott’s unlawful conduct.” Six days after that visit, ElectrifAi announced that the company was appointing female leaders to executive leadership positions. In advance of this story appearing and more than two weeks after India celebrated its 74th Independence Day on Aug. 15, ElectrifAi posted drawings of children of employees celebrating the day on its LinkedIn page.

“ElectrifAi’s infringement was and continues to be willful”
Kumar wouldn’t comment, but the whistleblowers allege that pleasing Scott was crucial to working at ElectrifAi. They also refer to his mercurial nature.

“To be honest, the atmosphere felt like Soviet Russia, where people who were praised one day would be replaced the next,” wrote a source in an email. “Things deteriorated further when entire lower/mid rung consultants were fired. Campus offers were also rescinded. After this, only sycophants who agreed with Ed were hired or were retained.”

While the employment discrimination lawsuit alleges many objectionable behaviors by Scott, it doesn’t address the efficacy or lack thereof of ElectrifAi’s PulmoAi product. Sitrick pushed back on many of the whistleblower’s allegations about the technology but did not address their central criticism: that ElectrifAi/Opera has historically no experience with medical imaging or with clinical diagnoses. 

The whistleblowers are a bit divided over why Scott and the senior leadership team at ElectrifAi are promoting a product they believe was built without real clinical expertise. Some say it’s a desperate ploy to garner revenue amid Electrifai’s falling business fortunes [which Sitrick denies] while others say it is an example of outright deception.

The latter perspective seems to have gotten some legs. On July 28, a Palo Alto, California company filed a breach of contract and trade secret misappropriation lawsuit against ElectrifAi.  The complaint states that Actian, a data integration, management and analytics company terminated a licensing contract with ElectrifAi in November 2019 because the latter failed to make payments and fees for the use of its products, among other reasons. Actian alleges that even after it terminated the licensing agreement, ElectrifAi is still using its products, thereby constituting an infringement of its software copyrights.

“ElectrifAi’s infringement was and continues to be willful and intentional,” the complaint declared.  

 Photo: ElectrifAi