FEDML Empowers On-Premise AI Innovation at DENSO

PALO ALTO, Calif.--(BUSINESS WIRE)--Apr 30, 2024--

FEDML, the company providing "Your Generative AI Platform at Scale," today announced its partnership with DENSO, one of the largest automotive suppliers in the world, to empower fully on-premise training, development, and deployment of AI models and applications via FEDML Nexus AI platform.

This press release features multimedia. View the full release here: https://www.businesswire.com/news/home/20240430680814/en/

As enterprises and organizations move fast toward bringing AI into their products and services, the need for privacy, security, full control, and ownership of the entire AI software stack becomes a critical requirement. This is especially true with the emergence of Generative AI models and applications, as data and AI models have become essential assets for any organization to obtain their competitive advantage.

Your Generative AI Platform at Scale – this is the mission that defines FEDML and its commitment to helping enterprises navigate the AI revolution with full ownership and control. By deploying FEDML Nexus AI platform on their own infrastructure (whether private cloud, on-premise servers, or hybrid), companies can provide their employees and customers with scalable, state-of-the-art GenAI capabilities, while giving them full control over their data, models, and computing resources.

Dr. Toru Hirano, Vice President of DENSO’s Silicon Valley Innovation Center, recently highlighted the complexities associated with training AI models across diverse geographic regions and internal business units. "Due to increasingly stringent data regulations and internal confidentiality policies, federated learning has emerged as an ideal solution by leveraging edge computing," Dr. Hirano stated. He further noted the limitations of existing methodologies: "Sometimes, even sharing local models with differential privacy and secure aggregation is prohibited, necessitating a purely on-premises solution." Through a partnership with DENSO, FEDML has successfully implemented such a solution, addressing the specific requirements of secure, decentralized data handling. This solution is being utilized by DENSO in such projects as creating ADAS features and developing AI applications designed to safeguard vulnerable road users (VRUs) in smart cities.

FEDML Nexus AI Key Features. FEDML Nexus is an all-in-one powerful AI platform that provides the foundation to build and commercialize any AI application easily, scalably, privately, and economically. The platform offers several key features:

According to Salman Avestimehr, CEO and Co-founder of FEDML, “In the last decade, many enterprises evolved into data-centric organizations. Now, with Generative AI advancing rapidly, these companies are transforming into AI powerhouses. FEDML Nexus AI facilitates this transition, offering full ownership and control. Our partnership with DENSO exemplifies how companies can maintain control over their data, models, and computing resources, leading the AI revolution.”

FEDML Nexus AI On-Premise Deployment. Deploying FEDML Nexus AI on-premise enables any enterprise to operate its own AI platform on its infrastructure, ensuring maximum privacy, ownership, and control. An engineer or developer simply needs to download the Nexus AI platform from FEDML’s private repository and install it on a node inside the enterprise’s private network. Once the Nexus AI web platform is online, enterprise users can login to the platform, register their private computing nodes, and execute different jobs, such as AI model training or deployment, to their private computing cluster.

FEDML Federate. One of the core components of the Nexus AI Platform is FEDML Federate, a state-of-the-art Federated Learning Operations (FLOps) platform for managing the end-to-end development lifecycle of Federated Learning jobs. FEDML Federate seamlessly orchestrates all operations with respect to the coordination, execution, and monitoring of a federated learning job. Through FEDML Federate, enterprises can federatedely train, deploy and monitor any AI model across their own private compute nodes (servers, devices, smartphones), enhancing their data privacy and security compliance. At the same time, through the advanced capabilities offered by FEDML Federate for easily experimenting with different federated learning algorithms and training policies, A/B model testing and group management features, enterprises can iterate faster and optimize their development costs.

In the figure 2, we provide a representative example of the execution flow of a federated learning job deployed on-premise through FEDML Federate within an enterprise’s private network. FEDML Federate will spawn and manage the required master and client processes at the registered computing nodes and provide advanced job, model and system monitoring capabilities through the on-premise deployed FEDML Nexus AI Platform.

About FEDML, Inc.

FEDML is your generative AI platform at scale to enable developers and enterprises to build and commercialize their own generative AI applications easily, scalably, and economically. Its flagship product, FEDML Nexus AI, provides unique features in enterprise AI platforms, model deployment, model serving, AI agent APIs, launching training/Inference jobs on serverless/decentralized GPU cloud, experimental tracking for distributed training, federated learning, security, and privacy. FEDML, Inc. was founded in February 2022. With over 5000 platform users from 500+ universities and 100+ enterprises, FEDML is enabling organizations of all sizes to build, deploy, and commercialize their own LLMs and Al agents.

Figure 2: FEDML Nexus AI Job Submission and Execution Pipeline. (Graphic: Business Wire)

Figure 1: FEDML Nexus AI On-Premise Deployment. (Graphic: Business Wire)

PORTLAND, Maine (AP) — An independent commission investigating the deadliest shooting in Maine history plans to explore allegations that police officers arriving without orders created “chaos” during the search for the gunman.

Chairman Daniel Wathen said commissioners intend to address some of the “disturbing allegations” in a report by the Portland Police Department special reaction team. But he said others may be outside the panel’s scope, including an allegation that deputies in an armored vehicle had been drinking before nearly crashing into another armored vehicle.

The panel reconvenes Friday to hear from witnesses on communications and coordination problems.

The Cumberland County tactical vehicle could have claimed lives if it had crashed into the Portland police vehicle, the report said, while noting that self-dispatching officers created the potential for more harm than good.

“I have never seen the amount of self dispatching, federal involvement with plain clothes and utter chaos with self dispatching in my career,” the tactical team leader, Nicholas Goodman, wrote in the partially redacted report, which The Associated Press obtained Tuesday through the state's Freedom of Access Act.

The Cumberland County sheriff refuted accusations that any of his deputies were intoxicated after the report indicated the armored vehicle skidded to a halt to avoid a crash that could've been fatal.

Both the Cumberland County Sheriff's Office and Portland Police Department tactical teams were responding to a location where the shooter's vehicle was abandoned by the Androscoggin River on the evening of Oct. 25, after the gunman, an Army reservist, killed 18 people and wounded 13 others at a bowling alley and a bar and grill in Lewiston.

The commission previously heard testimony from law enforcement officials about the chaotic hours after the shooting in which agencies mobilized for a search and police officers poured into the region.

The Portland report was especially critical of self-dispatching officers. The report suggested officers who arrived to help in plain clothes — “similar clothing to the suspect” — created a dangerous situation in which police officers could’ve exchanged fire with each other in a wooded area near the gunman’s abandoned vehicle. The gunman's body was found two days later at a nearby location where he died by suicide.

Tactical vehicles used by the Cumberland Sheriff’s Office and Portland Police Department apparently weren’t aware of each other’s presence.

The Portland team, which arrived first near the site of the gunman's vehicle, was attempting to keep police cruisers off a bridge where lights were transforming officers into targets when an armored vehicle approached from the other side of the bridge and skidded to a stop within 20 to 30 feet (6 to 9 meters) of the Portland vehicle, the report said.

In the report, Goodman wrote that the “smell of intoxicants” wafted from the tactical vehicle operated by Cumberland County tactical team members, who said they were assisting after attending a funeral.

Cumberland County Sheriff Kevin Joyce said in an earlier statement that any report of intoxicated officers should've been raised at the time — not six months after the tragedy — and he defended his deputies. He said in the statement that an internal investigation cleared the officers and that no one was determined to be intoxicated at the scene.

FILE - Crime scene tape still surrounds Schemengees Bar & Grille, Oct. 29, 2023, in Lewiston, Maine. An independent commission investigating the deadliest shooting in Maine history plans to take up accusations in a report that contended self-dispatching police officers created “chaos” during the search for the gunman, but may not address an allegation that deputies in an armored vehicle had been drinking before nearly crashing into another armored vehicle. (AP Photo/Matt York, File)