APC aimed to increase load flexibility and encourage customers to shift energy usage from peak hours to economy pricing hours during summer and winter months. The challenge was to achieve this without compromising customer comfort, which is crucial for maintaining customer participation in the program. The existing Time of Use (TOU) demand response (DR) program needed enhancements to optimize load shift while ensuring customers remained comfortable and satisfied.

Implementing new rate plans to align demand with supply is becoming critical to more and more utilities. However, new rate plans for business and residential customers can result in customer confusion and frustration—whether they are opt-in or opt-out. According to J.D. Power, utility customer price satisfaction drops by 10% when defaulted onto a TOU rate. A TOU rate transition can be stressful for both utility employees and customers. It can seem that customer needs are at odds with regulatory and internal requirements. Thankfully, utilities have gained experience helping customers through rate transitions. This experience identified best practices of the three e’s: empathize, educate, and empower.

Washington Gas aimed to drive energy awareness and help customers take control of their energy usage in Maryland and Virginia. They offered various programs, including rebates on high-efficiency natural gas equipment and Home Energy Conservation Kits. The challenge was to increase awareness and enrollment in these programs. To achieve this, Washington Gas partnered with Uplight to launch the Online Home Energy Profile, an online survey tool that assessed home energy usage and provided personalized recommendations. The goal was to drive program enrollment and encourage customers to adopt energy-saving measures.

Duke Energy, the largest utility in the United States, faced the challenge of enhancing its Behavioral Energy Efficiency (BEE) program to better serve its 7.5 million customers across multiple states. The primary goals were to support Duke Energy's brand positioning as a Trusted Energy Advisor, help residential customers save money, increase customer satisfaction, cross-promote energy products and services, and positively contribute to Duke Energy's financials. The program needed to be scalable, engaging, and capable of delivering personalized energy-saving insights to a diverse customer base. Additionally, Duke Energy aimed to reduce the opt-out rate and ensure that the program's communications were effective and well-received by customers.

Watase bootstrapped the company with money raised from business competitions and student loans. He wants to pump the market full of scholarship money “so that a million students can go to college”. His platform can do everything from vet grades and letters of recommendation to screen essays (although donors always have the final say). Scholars App awarded $20 million and processed over 100,000 applications in its second year alone. What is even more impressive, is that the average scholarship posted on Scholars App gets six times the qualified and complete applicants than they had before and this has encouraged donors to increase their scholarship funding. “About half of the donors actually give more money within eighteen months of signing up with our platform,” says Watase. It turns out that there are a lot of scholarship programs that can’t find qualified applicants. About 10% to 20% of scholarship monies aren’t awarded for this reason, Watase says. Scholars App doesn’t charge students fees and notes on its website “we background check every scholarship. We check that students will not have their data stolen and that they have a chance of actually receiving money.’’

Taking on the gargantuan task of assessing carbon sinks, Sylvera needed to accurately verify the performance of the projects they rate. This required precise tracking of land use and its evolution over time, particularly focusing on mangroves, which are crucial for absorbing more carbon than regular tropical forests. Sylvera knew that bringing in students and interns was one option to get the job done, but the company also knew how much additional interviewing, hiring, onboarding, training, and management that approach would take.

While Matterport focuses on delivering the highest value possible to its customers through the creation of cutting-edge technological solutions, it also needs to maximize the value of its extensive database and huge library of images. The company owns the largest spatial data library in the world (and it gets bigger every day). Its vast spatial data library enables the development of new neural network architectures and the ability to evaluate and optimize them against user behavior and real-world data in millions of situations. To create robust, highly accurate 3-D digital twins, it’s vital for Matterport to continuously optimize these neural networks and machine learning algorithms so that essential knowledge can be extracted from the growing volumes of data. Due to the sheer number of images captured, the spatial data company needed a partner that could provide image annotation and classification services for its database of 3-D panoramic scans. Instead of spending time on data labeling, the company wanted its in-house resources to focus on development.

Nearmap acquired proprietary roof geometry technology to expand its service offerings. The challenge was to accurately label high volumes of specialized roof facets in 3D content to operate at scale. Speed was essential for quick turnaround times, and the complex nature of the work demanded intensive training for data labeling teams. Additionally, Nearmap needed flexible support to handle seasonal demand fluctuations.

The first among aerial imagery providers, Nearmap offers both AI analysis and high-definition aerial images with accuracy on a commercial scale. Its geospatial data can help speed up and simplify many organizations’ workflows, from insurance underwriting to property appraisals, by processing virtual location data without the need for in-person inspections. Automating AI data sets requires up-to-date information and labor-intensive data labeling. When an AI component was first introduced to Nearmap’s offering, data labeling was done in-house, distracting the team from developing software. The Nearmap team quickly realized that to expand the business, they would need outside help. In addition to the sheer volume of images that need processing, a key challenge is the variety of imagery assets, and the different ways they must be labeled. Nearmap had to develop an in-house labeling tool, as no commercial tool exists that could meet their imaging requirements, which change frequently. Business growth meant finding a reliable partner that understood the complexities of what was required, could learn and work with the proprietary tooling, and ensured fast delivery of high-quality labeling work.

Ibotta faced a significant challenge in managing the receipt moderation process, which required matching items on a receipt with those selected in the app. The process needed to be both fast and accurate to ensure a high-quality user experience. While the company's proprietary OCR system handled most receipts, a substantial number still required human verification. During peak times, such as holiday shopping seasons, the workload became overwhelming, even requiring the CEO to assist in moderation. Ibotta needed a solution that could meet or exceed current performance benchmarks, improve accuracy by 10-15%, and increase efficiency by nearly 50%.

Chatbots often face limitations in answering diverse questions or managing various tasks, leading to rigid and frustrating user experiences. True Lark aimed to overcome these limitations by focusing on specific industries to provide a full range of necessary context. However, to achieve this, they needed labeled data to build their models, which required extensive time and effort. True Lark's internal team was unable to dedicate the necessary time to data annotation without neglecting other critical aspects of product development. This led them to seek external assistance from CloudFactory for data labeling operations.

Before partnering with CloudFactory, IC struggled to manage rapid growth because of the inability to process a high volume of complex information quickly from its many agency partners. The company faced limitations in the number of quotes it could offer due to the time-consuming process of gathering necessary information. This bottleneck hindered their ability to expand their product line and sell more insurance in a competitive market. The challenge was to find a solution that could efficiently handle the influx of data and streamline the quote generation process.

As important as recycling is to the sustainability of our planet, the challenge of doing it efficiently and cost-effectively has evaded the industry. While manual labor has been the mainstay, there are limitations to human perception that AI can overcome. There is immense potential for profitability by integrating this technology, however, it can only happen with high-quality annotated data being fed to the neural networks. Beginning as a small, cross-functional team, many of AMP Robotics’ team members had to share the load of data annotating. However, as the company began growing, the need for additional help became evident. AMP then took the step that many fast-expanding tech companies do and began hiring temp workers through an agency. “But we found that it was expensive, and the quality of the annotations just wasn’t there,” explains Ben Clint, AMP Robotics Data Acquisition Manager. “College students were brought in and trained, but the problem is that part-time, unfocused work is not the best thing to get really high-quality results.”

MAI needed the labeling work done by a consistent group of individuals who could log on remotely to the company’s own annotation tool which is custom-designed with machine learning components built-in. The work is critical to MAI’s efforts to stay ahead of the curve in providing AI-based image databases that enhance medical professionals’ understanding of health issues. One of its goals is to provide predictive advice based on tagging and analyzing images throughout a patient’s lifespan in order to enhance preventative care. MAI’s owner researched multiple companies looking for the best image tagging help. He was worried about crowdsourcing because he didn’t think the quality would be there. Some companies had preconceived ways of working with clients and were not interested in having staff work directly on his platform. Additionally, he needed workers used to dealing with images because each batch is unique in terms of what is being tagged.

Hummingbird Technologies faced the challenge of tagging and annotating vast amounts of data captured from drones and satellites to build accurate machine learning models for crop analytics. The process was highly domain-specific and time-consuming, requiring expertise in agronomy and remote sensing. The company needed a scalable solution to handle the increasing volume of data and to ensure the accuracy and reliability of their AI models, which are critical for providing actionable insights to farmers. Additionally, they had to continuously update their models to account for fluctuations in climate and other irregularities, which added another layer of complexity to their operations.

The IncucyteⓇ Live-Cell Analysis System automates live-cell imaging, producing a vast number of microscopic images that are beyond human capacity to analyze. Neural networks can handle this task, but they require a large, annotated dataset for training. Sartorius needed to identify and characterize individual cells within these images, which contain hundreds to millions of cells in various shapes and sizes. Although the IncucyteⓇ system has integrated software for automatic cell segmentation, it requires user training and bespoke analysis for different cell types. The neural network method, however, allows for a single analysis across a wide range of cell types without user training. Faced with the diversity and volume of cells needed for the training dataset, Sartorius realized they couldn't manage the annotation process alone.

Once SSC collects sensor data, it needs to be manually analyzed and then run through its machine learning algorithm. The company tried three options before turning to CloudFactory. First, SSC attempted to use on-site contract workers, but they were inconsistent even with the CTO there to train them. The crowdsourcing vendor chosen next wanted the work done in their tool, while the company had its own. None of the methods provided reliable results. Frustrated with these efforts, a company employee suggested CloudFactory as he had a positive experience with them in a previous position. The CTO found CloudFactory to be a perfect fit because of the dedicated workforce.

AI-enabled products that can record and monitor African wildlife come with their share of challenges. In addition to requiring massive amounts of training data, the diversity of the data must account for species, landscape, cultural relevance, and human influence. Unmanned aerial vehicles (UAVs) have proven to be a viable way to capture large amounts of data, however, these aerial surveys result in countless hours of video footage that can make finding value in the data collected challenging. If processed by humans alone, the work can prove to be mundane when there’s nothing of interest on the screen for hours on end. This is where machine learning proves useful, and the accuracy of the model depends on the accuracy of the data used to train the algorithm. To ensure the highest quality training data, Vulcan partnered with Sama, hiring a dedicated team of data annotators to put bounding boxes around key areas of interest in videos and images, and then pass the data back to Vulcan’s machine learning team to build various ML models.

Glassdoor, with its extensive database of over 8 million company reviews, salary reports, interview reviews, and questions, faced the challenge of maintaining the quality and reliability of its content. With a growing community of over 30 million users, ensuring that the information provided is accurate, relevant, and free from fraudulent activity became a significant concern. The platform needed a robust solution to handle content summarization, moderation, and verification tasks efficiently. Additionally, the need to filter out bogus information and augment data to enhance user experience was paramount. Glassdoor required a dedicated team capable of performing these tasks while also developing new training materials to improve the overall process.

The insurance provider needed to classify and extract information from tens of thousands of complex websites to automate answering questions for downstream applications.

CSET's data science team wanted to build NLP applications that classify scientific articles such as technical papers on virology. The team realized that manual labeling workflows would be impractical for the task.

Apple faced a significant challenge in supporting engineers who were dealing with contradictory or incomplete supervision data. This issue was critical as it impacted the accuracy and efficiency of their machine learning models. The traditional methods of data labeling were not only time-consuming but also prone to errors, which further complicated the problem. Apple needed a robust system that could handle these complexities while ensuring data privacy and reducing costs. The existing solutions in the market did not meet these requirements, prompting Apple to develop an in-house solution that could address these specific challenges effectively.

The bank needed an accurate way to tag companies in unstructured news text, link them to identifiers (e.g., stock tickers), and classify mentions by sentiment and other aspects. The existing solutions, including a black box vendor system and internal heuristic approaches, were not meeting the performance requirements. The bank required a more efficient and accurate method to handle the vast amount of unstructured data feeds and derive meaningful insights from them.

Rapidly changing sales goals make social media monitoring difficult to maintain. Intel faced challenges in keeping up with the dynamic nature of sales targets, which required constant monitoring and analysis of social media data. The traditional method of using crowdworkers for labeling data was not only costly but also time-consuming, leading to delays in actionable insights. This inefficiency hindered the ability of sales and marketing teams to respond swiftly to market changes and customer needs.

Building AI applications to extract entities requires high domain expertise and large amounts of labeled training data, which is expensive and time-consuming. The biotech company faced the challenge of processing a vast amount of clinical trial documents to extract critical chronic disease data. Traditional methods of manual labeling were not only slow but also costly, making it impractical for the scale required. The need for a more efficient and accurate solution was paramount to meet the demands of their research and development processes.

The bank needed to extract structured financial data from balance sheets and income statements (hOCR PDF) from private company financials. This task was challenging due to the unstructured nature of the data and the need for high accuracy in financial reporting. Traditional methods, such as manual data entry or rules-based systems, were time-consuming and prone to errors. The bank required a solution that could automate the extraction process, improve accuracy, and handle the variability in document formats and data presentation.

The time-consuming process of manually labeling high-cardinality training data blocked Pixability from expanding their NLP capabilities.

Labeling training data for triaging models in medical imaging is a time-consuming process, often requiring person-months to person-years of radiologist time. This manual labeling is not only labor-intensive but also prone to human error, which can affect the accuracy and reliability of the models. The challenge was to find a more efficient and accurate method to label large datasets of medical images, which are crucial for developing and training machine learning models for disease diagnosis and patient monitoring.

New innovative concepts for mobility, such as highly automated or autonomous driving, place enormous demands on the safety and reliability of technical systems. The efficient development of particularly reliable control systems for autonomous driving up to SAE Level 5 requires the use of suitable technologies. Conventional control-based approaches compete with the capabilities of trained neural networks, which are particularly suitable for processing vast amounts of data from high-resolution sensors. The first step is to identify potential application areas for AI along the entire processing chain from perception to situation analysis and behavior planning. Promising methods from machine learning must be evaluated, focusing on multimodal perception, i.e., the environment perception of a vehicle with merged data from video, radar, and lidar sensors.

With a spotlight on public sector organizations, Camden sought to add an additional layer of security that ensured information would be secured and protected from unauthorized access and usage. Camden also wanted a tool to boost efficiencies by improving the utilization of existing assets rather than buying more of them. Camden wanted to increase its overall productivity: IT managers did not have a streamlined solution for mass deployment over large and complex estates.