Participated · Series A · Apr 2022

Synthesis AI offers a cloud platform that produces labeled synthetic image data using AI, procedural generation and VFX rendering to train computer-vision models. The platform can generate diverse digital humans and customizable avatars, and the company says it has delivered over 10 million labeled images and produced datasets such as roughly 100,000 synthetic people for customers. Customers span AR/VR, metaverse, robotics, consumer devices and automotive driver/occupant sensing, including Fortune 50 clients. Synthesis plans to launch new products targeting photo enhancement, teleconferencing, smart homes and smart assistants while expanding research into mixed real and synthetic data. The company emphasizes ethical development—arguing synthetic data can reduce bias and preserve privacy—while acknowledging limitations of synthetic-only training. Synthesis currently has 20 employees and plans to scale to 50 by the end of the year. Synthesis AI is a San Francisco-based provider of synthetic data-as-a-service that builds synthetic image datasets to train computer vision models. The company uses a proprietary combination of generative neural networks and cinematic CGI pipelines to programmatically create vast amounts of labeled image data at speed. Its platform scales in the cloud from research and development phases with small amounts of data to production requirements of terabytes of data. Synthesis AI has launched FaceAPI, an on-demand product that can generate millions of perfectly labeled human images; FaceAPI is generally available and is currently in use by major technology and handset manufacturers. The company raised $4.5M in additional funding and intends to use the proceeds to add to its R&D teams. The round was led by existing investors including Bee Partners, PJC, iRobot Ventures, Swift Ventures, Boom Capital, Kubera VC and Leta Capital.

Participated · Series C · Apr 2022

Sense builds grid-edge intelligence software that runs directly on modern smart meters, transforming high-resolution waveform data into real-time visibility for utilities and actionable energy insights for homeowners. Utilities use the platform for consumer engagement, load visibility, demand forecasting, and fault detection, while end users access a mobile app to monitor and manage household energy consumption. By collaborating with meter manufacturers, Sense integrates its algorithms into next-generation advanced metering infrastructure (AMI). The company projects deployment of its software on tens of millions of U.S. meters over the next decade, dramatically expanding its data footprint. This embedded, AI-driven approach promises faster responses to grid issues and better alignment of household demand with grid availability. Cambridge-based Sense positions itself as a key enabler of a more resilient, affordable, and sustainable electric grid.

Led · Seed · Jan 2022

Labrador Systems develops Retriever, a domestic assistance robot that resembles a robotic bar cart and is aimed at elderly users and people with limited mobility. Retriever can carry up to 25 pounds, uses a retractable tray to move objects from counters and shelves, and includes additional storage and a phone-charging port. The system will feature Alexa voice control and the company says it plans to offer a bespoke refrigerator accessory. Labrador began rolling prototypes to select users and is running testimonial tests; after a beta period it plans commercial sales in the second half of next year at multiple price points. Early-adopter pricing is $1,500 up front plus a monthly service fee of $99–$149 depending on financing. The company’s backers include Alexa Fund, SOSV/HAX, iRobot and the National Science Foundation. Labrador Systems develops assistive robots for the home, focusing on practical robotics applications that can address immediate real-world needs. The company is building home-assistive robotic solutions and is positioning its products to handle household and non-patient tasks. Labrador recently closed a $2 million seed round to fund development of its assistive-robotics offerings. The company is based in Los Angeles and is led by cofounder and CEO Mike Dooley. Dooley previously served as VP of Product and Business Development at iRobot. Labrador was scheduled to appear at TC Sessions: Robotics+AI to discuss assistive robotics on March 3.

Participated · Series C · Aug 2021

Osaro develops advanced robotic automation systems that combine machine learning–based object recognition with adaptive control software to enable logistics operators to scale e-commerce order fulfillment, inventory induction and extraction, sortation, and singulation. Its robotic piece-picking solution is optimized for high-SKU inventories, complex packaging, and fragile items, and integrates with ASRS, AMRs, conveyor systems, and other materials handling equipment. The systems have been used by the world’s top retailers, system integrators, and third-party logistics companies. Led by CEO Derik Pridmore, the company plans to extend its solution and support offerings and to leverage its machine learning platform into new automation applications. Osaro will also use the funds to accelerate global growth through market expansion and hiring. Financially, the company raised $30M in a Series C round, bringing total funding to date to $67.5M. OSARO develops machine-learning software for industrial automation, with a focus on robotic piece-picking for e-commerce order fulfillment and intralogistics. The company offers a software-only, vision-first approach and is the only full-stack robotics firm in the article that provides a vision-only solution tightly integrated with major industrial arms and grippers. OSARO's software combines advanced object-recognition models with motion control to automate picking, packing, and related warehouse tasks across industries including electronics, apparel, grocery, and pharmaceuticals. In 2020 the company deployed 18 software systems, grew its team by over 20% (adding ten engineers), and launched a Tokyo subsidiary, OSARO GK. OSARO also collaborated with Innotech to produce a mobile “Piece Picking Station” marketed to logistics and factory automation customers. Performance evaluations cited in the article report a 99% successful pick rate, a 99.81% successful placement rate, and the highest accuracy across 1,400 unseen and adversarial objects. Financially, the company has raised a total of $37 million from investors on four continents. OSARO builds hardware-agnostic machine learning software for industrial automation, with flagship products OSARO Pick (robotic piece-picking) and OSARO Vision (perception and control). Its software automates stationary picking stations in "goods to robots" distribution centers and integrates with a wide range of robotics hardware. The company emphasizes deep learning algorithms that generalize picking tasks with minimal training data and no SKU registration to enable quick, scalable deployments. OSARO operates a cloud-based data and GPU infrastructure to support industrial-scale, fault-tolerant robotic deployments and began deploying OSARO Pick in Japan in early 2018. Planned uses of new capital include talent acquisition, international deployments, and advancing the OSARO Pick and OSARO Vision product lines, with targeted deployment growth in North America, Australia, Korea, China, Japan, and Germany. The company also announced a senior engineering hire (Kevin Pope) to support global scaling. Osaro is a nine-person, San Francisco-based startup developing advanced machine learning (deep reinforcement learning) that combines perception with decision-making. Its core product is a machine-intelligence operating system for robots that learns via trial and error and will be licensed to customers. The company is exploring pilot deployments in industrial robotics and is working with firms that have critical needs for more flexible automation; it currently has no commercial customers. Osaro aims to enable lower-skilled technicians to set up and train robots over time and ultimately to improve agents remotely by learning across deployed robots. The startup raised $3.3 million in seed funding to take its technology to market. Investors in the seed include Scott Banister, Jerry Yang’s AME Cloud Ventures and Peter Thiel.

Participated · Seed · Dec 2020

Hasty develops annotation tools for vision AI practitioners that are supported by a community of machine learning engineers, data scientists and software developers. Its annotation tool labels data quickly through self-learning AI-assistants and uses agile machine learning to provide rapid feedback so engineers can validate and adapt models as they work. The company is led by co-founder and CEO Tristan Rouillard. Hasty plans to use new funding to continue accelerating product development. It also intends to expand its customer base across Europe and North America. The company is based in Berlin, Germany.