Our verdict
DataRoot Labs is the firm startups and product teams hire to build custom machine learning, LLM agents, and robotics systems, with an engineering team in Kyiv, Ukraine. It does not publish rates; outside listings put it at about $50 to $99 an hour, with projects starting around $10,000 and most reviewed projects costing $50,000 to $199,999. At about $50 to $99 an hour, it is in the lower half of the firms reviewed here that list hourly rates, level with 10Clouds, Arionkoder, BairesDev, and HatchWorks AI, and higher than 10Pearls at about $25 to $49. Other firms reviewed here that build custom AI agents and machine learning systems include InData Labs of Nicosia and Neurons Lab of London, though Neurons Lab works only in financial services. It is a strong choice for a company that wants a research-minded team to co-build the technical core of an AI product, but it is a small firm and its case studies seldom give measured results.
The right buyer
- Startups and product companies that want an AI team to co-build a core feature. Its homepage shows the logos of Outrider, an autonomous yard truck maker, and Moxie, a social robot for children whose maker, Embodied, announced in December 2024 that it was closing (homepage; Ars Technica report).
- Teams building LLM agents over private data with human review. Its cases include an incident classifier that safety specialists approve (case study), a financial document copilot that links answers to page and table references (case study), and an HR assistant covering more than 20 countries (case study).
- Companies with robotics, computer vision, or sensor data. It built a simulated robot inspection prototype in Isaac Sim (case study) and audio and motion models for a pet wearable that handled more than 10,000 concurrent users in load testing (case study).
- Businesses with heavy data flows, such as a vehicle telematics service. DataRoot Labs built the ingestion and event processing on Scala, Kafka, Spark, and Cassandra (case study).
Reasons to look elsewhere
- It publishes no rates or package prices; it describes three engagement models, a research partnership, product co-building, and a dedicated team, without costs (homepage).
- Most case studies describe outcomes without figures, for example a classification workflow that is “more consistent” (case study).
- It is small. Its LinkedIn page describes a team of 50 or more, lists a company size of 11 to 50, and shows 24 employee profiles (LinkedIn).
- Delivery runs from Kyiv, and the homepage lists no other engineering office. One client reviewer asked for a London office (see the client reviews in the “What clients say” section).
DataRoot Labs, At A High Level
We read DataRoot Labs’ homepage, LinkedIn page, case studies, and DesignRush profile to establish who the firm is, how large it is, and what it has built. DataRoot Labs is an AI research and development company in Kyiv, Ukraine, that builds custom machine learning, LLM agent, and robotics systems for companies from Los Angeles to Tel Aviv. It also runs DataRoot University, a free trainee program in data science and engineering that its LinkedIn page says more than 6,000 people have completed since 2017 (LinkedIn).
- Founded in 2016 according to its LinkedIn page and its DesignRush overview, though a DesignRush biography says CEO Max Frolov started the company in 2014 while completing his master’s degree. Headquartered in Kyiv, with a US address in Wilmington, Delaware (LinkedIn; DesignRush profile).
- Describes a Ukraine-based team of 50 or more professionals and more than 70 AI projects in over 10 years (LinkedIn).
- Organizes its work into three labs: an Agentic Lab for agents, retrieval, and evaluation; a Deep Learning Lab for model training, reinforcement learning, and robotics; and an Engineering Lab for AI products, APIs, and data systems (homepage).
- Offers research partnerships, product co-building, and dedicated R&D teams, with scope, responsibilities, and IP agreed before work starts (homepage).
- Says it helps companies such as IBM, Noom, and Cognyte co-build AI components into their core (LinkedIn), and DesignRush lists clients including Django Stars, OLX, Databand, and The Meet Group (DesignRush profile).
- Works with Kyiv Polytechnic Institute on programs that bring AI into technical disciplines, under its DRL AI Space initiative (homepage).
Its documented client work includes:
- A safety incident classification service for a manufacturing safety team, built with LangGraph and Azure OpenAI, with safety specialists approving the output (case study).
- A financial document copilot that extracts findings from filings, memoranda, and earnings recordings and returns them with page, paragraph, or table references (case study).
- A global people operations assistant with role-based access and PII masking, covering policy lookup, document drafting, and HR system work in more than 20 countries (case study).
- A survey intelligence agent for a multinational manufacturer that analyzes more than 35,000 survey responses and produces reports on request by text or voice (case study).
- Audio and motion models for a connected pet collar, served through TensorRT and Triton, and tested with more than 10,000 concurrent users (case study).
- A distributed vehicle-data platform for a large telematics service and a learning and candidate matching platform for a technical talent marketplace (telematics case study; talent platform case study).
Scores by criterion
We scored DataRoot Labs on eight weighted criteria from its case studies, lab pages, client lists, and client reviews, to show which parts of its work are well documented and which are not. Commercial clarity scores lowest because DataRoot Labs publishes no price list.
| Criterion | Weight | Score | Basis |
|---|---|---|---|
| Production track record | 20% | 7.0 | Eight published cases across agents, models, and data systems, plus more than 70 projects reported; most outcomes are described without figures. |
| Technical depth | 15% | 8.0 | Three labs covering agent harnesses and evaluation, reinforcement learning and robotics, and high-load engineering, with a published toolbox. |
| Industry depth | 10% | 6.5 | Cross-industry work in HR, manufacturing, finance, logistics, and consumer devices, with no single sector focus. |
| Delivery model and handoff | 15% | 7.0 | Research, co-build, and dedicated team models with scope and IP agreed first; client reviews mention remote delivery. |
| Commercial clarity | 10% | 5.5 | No rates published; the three engagement models are explained but not priced. |
| Platform independence | 10% | 7.5 | Toolbox spans OpenAI, Anthropic, Google, Azure, and AWS services and open-source frameworks, with no single vendor focus. |
| Stability and scale | 10% | 6.0 | Operating since 2016 with about 50 people, privately held, delivering from Kyiv. |
| Independent evidence | 10% | 7.0 | Written client reviews from Django Stars, Pressmaster, and Dentsu Ukraine, and selected work for Outrider; no partner certifications listed. |
| Overall | 100% | 6.9 | Weighted average |
Where DataRoot Labs fits and where it does not
We rated DataRoot Labs against four jobs drawn from its case studies and lab pages, to see which ones its published work supports: a product co-build, an LLM agent project, a robotics or sensor project, and a large enterprise rollout.
| Need | Fit | Why |
|---|---|---|
| Co-building the technical core of an AI product with a startup team | Strong | Product co-building and dedicated team models, with reviews from Django Stars and Pressmaster. |
| LLM agents over private data with human review | Strong | Four agent cases with approval steps, source references, and access controls. |
| Robotics, reinforcement learning, and sensor models | Good | Simulation prototype in Isaac Sim, a pet wearable model stack, and a Deep Learning Lab built around robotics. |
| Large enterprise rollouts with change management | Weak | A team of about 50 and no published rollout or training practice. |
Buyers who want a Kyiv team for a consumer or fintech app can also look at Uptech, a product development firm with a lower hourly band of about $25 to $49.
What clients say about DataRoot Labs
Written reviews from the CEO of Django Stars, a co-founder of Pressmaster, and the CTO of Dentsu Ukraine give outside feedback on DataRoot Labs.
“Their level of AI expertise is incredible.”
“It didn’t feel like one of those teams that just stitches together APIs based on documentation. They had a real understanding of how the major models work beyond what’s publicly available.”
“DataRoot Labs proved to be an exceptional partner, demonstrating a high level of professionalism, expertise, and proactivity throughout the entire project.”
One client asked for a closer presence.
“Please open an office in London, it’d be a joy to work even closer. Even though you do a superb job remotely.”
Who leads DataRoot Labs
DesignRush lists four co-founders: Max Frolov as CEO, Yuliya Sychikova as COO, Ivan Didur as CTO, and Vasyl Koida as head of data (DesignRush profile).
Sources (12)
- DataRoot Labs: homepage
- DataRoot Labs on LinkedIn
- DesignRush: DataRoot Labs profile
- DataRoot Labs: safety incident classification case
- DataRoot Labs: financial document copilot case
- DataRoot Labs: global people operations assistant case
- DataRoot Labs: survey intelligence agent case
- DataRoot Labs: pet wearable case
- DataRoot Labs: robotic inspection in simulation case
- DataRoot Labs: telematics platform case
- DataRoot Labs: learning and candidate matching case
- Ars Technica: Startup will brick $800 emotional support robot for kids without refunds (December 10, 2024)