We received a lot of good reviews for the Top 6 AI Companies to Start Your ML Project article. Our readers were genuinely interested in how Artificial Intelligence developers from different parts of the world present themselves and communicate with potential clients. It’s no wonder because according to Google Trends the popularity of Machine Learning development in 2019 is increasing:
That’s why we prepared for you some more top Machine Learning companies. What expertise we have to evaluate them?
Here at SPD Group, we offer Machine Learning and AI development including Predictive Maintenance, Predictive Analytics, Virtual Personal Assistant, Cloud-based AI models, and Raw data management. We developed a predictive maintenance solution for 40 000+ payment terminals and NLP-based data clusterization solution for financial data and software company. If you are interested in Machine Learning development, have any questions and suggestions, or you want to start your ML project, feel free to contact us.
Like the previous time, we had asked for help from our own experts: Head of RnD Department Dmytro Kuriatnyk, ML Tech Lead Oleksandr Melnyk and Project Manager Olena Kovalenko. They are a part of the SPD Group and have practical experience in this field. We collected as much information as we could and assessed companies by the following criteria: Response Time, Expertise, Communications, Management, Types of Engagement and Pricing, Maintenance and Scaling. Each point is evaluated on the scale from 1 to 5, and that’s how we form the overall average rating that is displayed with the name of the company.
So here are six more of the best Machine Learning companies!
ML Company from the Ozark Mountains, USA – 3,8
This software development consultancy is from the Ozark mountains in Arkansas USA. Since 2014 they worked with a variety of companies ranging from MondoBrain, Airbus France and Peugeot, to CBRE. They respond in two hours, and you will be able to talk with the sales representative that will consult technical guys for help.
Dealing with the Machine Learning project they use Python stack. PyTorch/Tensorflow, SciKit learn, along with some proprietary tools.
They have their only office in Fayetteville, Arkansas and previously had effective collaborations with companies in France, California and other locations. They primarily use Slack, and occasional conference calls and emails. Every situation is unique, sometimes they can travel to the client, and the client can travel to their office, if necessary.
The company is a strong advocate of Agile Scrum methodology. They start every engagement with a week-long Ignition process where they learn about your company, create a roadmap and define core objectives. The Ignition process sets realistic expectations for the project, and that’s exactly what helps this company to stay on budget and meet the deadlines. It’s crucial because you pay for time, working with this company. Only after this the development start, where you will have one or two dedicated engineers as well as a full-time product owner. You will get access to the team during the entire development process.
Engagement models and pricing
The following models are offered:
- Ignition includes Roadmap Development, SWOT Analysis, Critical Path Analysis, User Experience Wireframing, Architecture & Infrastructure Plan, Product Backlog, Cost Estimates for Development & Operations.
- 8 Week Micro project. You will get a Product Owner, dedicated software engineer or senior software engineer.
- 12 Week Engagement. The same as 8 Week but longer and options for Staff Augmentation.
- They also have options for 6 to 36 months of engagement.
Payments are accepted via ACH transfer or check. They take two weeks’ payment upfront and offer net 15 terms with monthly billing.
Maintenance and Scaling
They provide Mission Control services following the development of the project to ensure long-term stability. The door is always open for future development with one-week sprints.
- Management and communications are good, a standard for the industry at the moment
- Ignition helps developers to get a better understanding of the needs of a client and the viability of the project
- Engagement models and pricing are clear and reasonable
- The focus on the small teams of developers is a little confusing
- The scalability is unclear
- This company claims that they have 15% focus on Artificial Intelligence development, however, during the conversation they didn’t provide any use cases or proofs of the previous projects in the Machine Learning field. On their website, there is also no information on the experience
ML Company from Minsk – 4
The mature software development company from Minsk, Belarus with almost 10 years on the market. They focus on designing and delivering new digital products and services for businesses across all industries and verticals. Co-CEO/Co-Founder replied in 50 minutes.
Their technical stack includes Python, SciPy, Apache Hadoop, Apache Spark, SparkML, Apache Kafka, PyTorch, Keras, TensorFlow, ElasticSearch. You can also count experience with ML Platforms: Azure Cognitive services, Google Vision and Amazon Rekognition. They claim to be working on a cropping algorithm for photo printing lab as a related project.
They are working with clients in the US and UK. Skype, Slack, and Emails are being used for communication. Group calls at least twice a week are recommended, but it could be more frequent if the client asks. For tasks and bugs management they use TargetProcess or Jira, Dropbox Paper or Google docs for collaborative document creation. Business trips are also possible, but expenses are covered by a client.
The team usually consists of Project Manager, developers, sometimes QA and UX/UI experts if it’s necessary. In most cases, PM is the main communicator, but it can be discussed. You will have the opportunity to talk to the developers directly. Developers for the project are changed only when there is a specific reason. As far as methodologies go, they use Agile, Scrum or Kanban.
The development process is transparent. The client gets updated on the statuses regularly. When there is some problem making a budget or a deadline, the clients will know it as soon as possible.
Engagement Models and Pricing
Three models are being offered:
- Time and Material with a clear hourly rate
- Fixed Cost Model for the projects with well-documented requirements
- A dedicated team with minimum requirements of 2+ team members for 6+ months
All payments go via a bank transfer, usually, the payment terms are NET 15.
Maintenance and Scaling
Scope change is possible during the development, PM will help to estimate cost increase. The maintenance process depends on a particular case: sometimes the main team is reduced to support the project, sometimes the SLA needs to be signed for a certain amount of hours every week or month.
- Communications and Management are looking great, all industry standards are met, and experts are there
- Maintenance and Scaling are good
- There is no information on prices
- They are lacking information on practical ML experience. On the official website, there is no mentioning of the development of AI/ML solutions from scratch. The only ML relevant experience in the development of chatbots on Alexa and building electronic device recognition based on smart algorithms for IoT projects.
ML Company from Kyiv – 2,5
Long-term technology partner for business with the hundreds of delivered projects, as they say. CEO replied in 53 minutes.
They state that Natural Language Processing, Computer Vision and Personalization in E-Commerce are the three strongest areas in their ML Expertise. Tensorflow and PyTorch among some other ML development toolkits are being used.
There only one office in Kyiv. This company is willing to communicate remotely or directly in their office. All travel expenses are on the client.
Their typical process looks like this:
- Investigation phase – analysis of current data and what can be collected, what are target goals, and what approach can be applied to achieve goals.
- Model MVP development – developing the first approach of the model and testing it on the real data
- Production model development
- Tuning and future development
Unfortunately, they can’t say more about the management process without personal meetings or Skype calls and signing NDA.
Engagement Models and Pricing
There is no information available.
Maintenance and Scaling
They claim that near 40% of the projects they develop are sustainably growing and supported after release. However, there are no details on how they achieve this.
- Response time is the main advantage of this company
- Management and Communications are decent but could be more detailed
- It is strange, that management process will be explained in detail after signing NDA, as well as Engagement Models, Pricing, Maintenance and Scaling
- They use PHP and Python + Vue, but AI/ML development is not the primary focus. Their website contains a big amount of basic Machine Learning information. No clear practical expertise in AI field or use cases.
ML Company from USA and India – 4,3
Full-service software partner specialized in architecting, designing & developing tech solutions by binding Web, IoT, Analytics, AI and Cloud technologies. Their global clients include organizations right from Enterprise, Midsize to startups. They replied in 2 hours and 15 minutes.
This company uses Node.js as the development platform for Machine Learning and AI capabilities. They have experience developing ML and AIalgorithmsm in Edutech, Retail and E-Commerce sectors. In Edutech they implemented Machine learning and AI using VARK learning methodology basically system tracks student learning behavior, learning pattern and automatically change the learning curve of the course.
They do daily standup meetings of 20-30 minutes via Slack or Skype. Jira, Basecamp or Monday.com are used for project management. They prefer Jira because it follows the Agile/Scrum methodology which allows them to manage the entire project. By managing the entire project they mean that the project will be divided into use cases/backlogs and by combining multiple backlogs they will create Sprints, basically deliverable which you can expect every week, or in 15 days depending on how big the project is. The company offers two travel options, depending on the type of engagement. In the first one, the traveling expenses are covered by the client, and in the other, they are split in a certain percentage.
They have a full-fledged development house situated in the central part of India & Marketing/Front desk office situated in Reno, Nevada in the USA.
The team is always formed based on the project requirements, which are established after multiple workshop meetings with clients. After the meeting is complete, they finalize the tech stack for any given requirement depending on business logic and scalability. Once team is selected then contract with individual team member is signed which binds them to be the part of the project till the end, and in case team members leaves then another team member is added, leaving team member passes the entire information to new team member and trains new team member till leaving date which comes under their notice period of 45 days. In these 45 days, entire work is handed over to new team member and multiple buddy jacking sessions are held. They take on the responsibility to replace the team member in case of non-performance and leaving, and they make sure to add a new team member based on the required skill set. You will have access to the whole team (from bottom to top of the hierarchy).
This company admits deadline or budget fail, claiming a 95% completion rate for their projects. The most common reason for failure is a requirement change, but there are also other multiple variables. To avoid such scenarios they started following Agile/Scrum development methodology, which sets clear goals in terms of the deliverable at the end of each sprint, and if there are any changes required apart of set project scope then it is clear to the client and to them and then deadlines and budgets are set accordingly. If they fail to meet the new deadline then, in that case, the X percent discount will be offered, which will be set between the client and their team of next sprint delivery.
Engagement Models and Pricing
After the project requirements are established and identified, the team is formed. It can include System Architect, Team Leaders, SR & JR Backend and Frontend engineers, Scrum Master or Quality Analyst. The team size may vary depending on project size and complexity. The hourly rates are clear. Once they obtain requirement understanding clear, which obviously will happen after multiple workshop meetings then they will give you information on expertise which is required, a number of team members and the monthly cost of the team. The company can set a monthly payment cycle, you will only have to send payment once in a month, deliverable will be set accordingly.
Maintenance and Scaling
Any changes during the development process are analyzed and if changes are required in adding a team member or tech stack, then it can happen on the go. Before any change is made they take approval from the client by notifying the changes they are making based on an addon to requirement. Once the project is complete and delivered, they set 2-3 members from the project development team as support staff for a client, support staff team size varies depending on the size of the project. Free support of developed modules until 3 months after development is guaranteed.
- Splitting travels expenses, which is unfortunately uncommon
- Guaranteed 3-month support of the project
- Knowledge transfer, discounts, access to the team – overall management process is great
- The lack of information on actual expertise. There is only one AI/ML business case in the answer and none on the website. Using Node.JS for AI/ML solutions is untrivial, in the aforementioned use case, the knowledge transfer process could be difficult
ML Company from San Francisco, California – 4,4
For a decade this company been building apps for iOS/Android devices and working on associated server logic. They have been working on dozens of apps being used by millions of people around the world. The response time is 4 hours.
They have worked quite a bit with TensorFlow and PyTorch for ML backends. Keras is usually the front end layer they bundle on top of TensorFlow. For image datasets, they use a software called LabelBox which allows them to scale up data collection and tagging across distributed teams. For numerical data, it is a bit more complicated process of harvesting existing datasets usually sourced through amplitude and more recently segment. But it is very dependant on the kind of data they are harvesting. This company gave a few use cases, showcasing certain technologies they use.
They use Slack and Email for communication. There is an option to add you to their team chat, where you can follow along with day-to-day progress. Each week there will be a detailed report from the team on the progress during the week, and the plans for the next one. Google Hangouts and Zoom are being used for meetings with video calls. You would be welcome to join those meetings, but they would want to have them from 10 am – 6 pm EST so the employees can maintain a normal work-life balance. After meetings, they typically have the team lead post their notes and/or action items in Slack or an email that you would be able to see. The team is located in one office in Columbia, Missouri (Central US).
Typically, they like to form teams with 1 part-time technical/project lead that is more senior and 1-2 full-time developers working with them. You will have access to the full team so you can talk 1:1 with developers when needed, but for the most part, the communication through the project lead is preferred, so they can keep things organized and on track and allow the developers to focus on their work. They admit failing deadlines only due to requirement change.
Engagement Models and Pricing
They would like to bill a flat hourly rate and set pricing expectations upfront. The monthly cap is set so you know the absolute maximum you are going to be spending month to month. Fixed bids are avoided since it is in direct competition to agile workflow and does not account to scope changes that happen as the project progresses. Before the project starts they do a thorough outline of task timelines and make solid cost estimates. The invoice comes once a month through QuickBooks online. It is a simple setup and allows you one-click payments.
Maintenance and Scaling
This company is ready to scale your project, and even offer a discount for longer terms. Post-release support can go two ways:
- They help you interview and hire your own full-time engineer and work to train them up on the stack. Then they remain on the project and work a set number of hours to help out when needed.
- They define the scope of work and continue improving the project.
- While the client needs to adjust to the team’s schedule, fixed working hours will lead to better performance of the development team
- The help from the company on interviewing and hiring employees for the maintenance phase
- Their response was professional and right on the point. The experience and expertise in building AI/ML solutions are very clear. The technology they use is modern and popular. Even more than that, they have an active blog on Medium
- No warranty period
Another ML Company from Minsk – 4,5
Custom software development agency, located in Belarus. Their clients are ambitious startups with big ideas, and established enterprises who are looking for advanced solutions. They help startups to build their big things and also help to get corporate software products from the stage of concept to production. The COO replied in 4 hours and 40 minutes.
In most projects they use PyTorch, but some models were implemented using Tensorflow. They had a few projects with real-time computer vision involved. It was real-time people face recognition/features extraction, people tracking, posture detection, hands gestures. As for regression/prediction tasks they’ve implemented prediction of sales and mobile user behavior prediction (purchase) for a mobile company. The company had an accuracy of 82% for churn prediction and this allowed to increase conversion rate up to +32% on specific funnels. Implementation was done on PyTorch.
They have huge experience working remotely. The company uses Slack for instant communication, where everyone on the project is added to a dedicated workspace. They have a rule that any direct question should be answered within 10 minutes. For voice and video communication, Hangouts\Zoom\Skype is being used. Youtrack is proposed for tickets\sprints management but you can suggest other systems. They claim to reach to you through the day in case of uncertainty or questions, and there will be demos of the project at the end of each sprint (generally each week or biweekly). You can also attend daily standup meetings, or they can make them together with your team if needed. All traveling expenses are covered by the client. Their office is located in one place in Minsk, Belarus.
The first step is preparing technical design – deciding on a tech stack and approaches, after that they will know which engineers will fit best. The standard project team looks like this:
- Developer (1 or more)
- Project Manager
- Quality Assurance Engineer
There is also a possibility to add additional roles:
- User Interface Designer
- Video Operator
Every team member should be approved by the client. When they assign engineers on a project they guarantee that they will stay on it until the very end, unless there is an emergency happened. You will be able to speak to everyone on the project.
They claim to meet the time and budget deadlines if the scope stays the same. Usually, they deliver once a week/once per 2 weeks and provide a customer with a visual changelog so you can visually track progress.
Engagement Models and Pricing
They work under the T&M model and bill hourly for project-based work and weekly if they provide dedicated developers to work under customer management. After signing the contract, they ask for an advance payment for 2 weeks. Then they use an invoice with NET15 payment terms every 2 weeks. But these conditions could change via agreement form both sides. They can work under a direct contract with the EU company, based in Estonia. This company can use almost any payment processor (Transferwise is preferred).
Maintenance and Scaling
You can change the scope of the project during the development process. Speaking of post-release support, they gave an example for production projects 24 hours response time on critical issues. They claim to fix the production app really quickly.
- The 10-minute rule for answering a direct question is a great feature
- Payment models are flexible, and everything about Engagement Models and Pricing is very clear
- ML development expertise is very good. The big plus is that they convert the accuracy of the model into KPI. This company has a blog on AI/ML, and the overall impression of their proficiency is very positive.
- The availability of the guaranteed period is unclear
Conclusion: How to Choose Your Top ML Company?
In the first article, it was concluded that trustworthy Artificial Intelligence and Machine Learning companies must prove their expertise and experience in this field, transparency in the prices and models, access to the development team, a variety of communication tools, openness in all processes and providing support periods. It still stands true. This time we have a few companies on our list with low scores. The main reason for this is a lack of information after direct questions and searching the websites and social media of particular companies. The lesson there for the companies using Machine Learning is to provide the fullest amount of information possible, to make a good impression. If you want to learn more about the companies in the article feel free to contact us, and we will provide more detail.
Machine Learning Consulting Companies
Artificial Intelligence and Machine Learning development can be very expensive and take a lot of time. Sometimes to get the most of the AI solution you need to transform some areas of your business. To make sure this is exactly what you need, you can get help from machine learning consulting companies. They have the typical projects of the ML implementation for different industries, they can give you honest numbers on AI capabilities for your particular case and what potential revenue it could bring. SPD Group is also providing consulting services, and if you are questioning the decision to go in the ML direction, we can help you to clear things out and build the most effective strategy. ML development is the part of USA’s National Artificial Intelligence Research and Development Strategic Plan, so the entire countries are considering ML solutions as well as you do!