Datacurve
Datacurve is a San Francisco-based AI data infrastructure company founded in 2024 that builds high-quality training datasets, reinforcement learning environments, and evaluation benchmarks, with initial focus on long-horizon software engineering tasks for frontier AI labs.
- Company typePrivate
- Founded2024
- HeadquartersSan Francisco, United States
- Headcount11–50
- GTM typeB2B
- OfferingSoftware
What Datacurve does
Datacurve is an AI data infrastructure company founded in 2024 in San Francisco that builds tools and datasets used to train and evaluate frontier artificial intelligence systems, with an initial focus on software engineering tasks. Its product portfolio centers on high-quality, long-horizon coding data and evaluation: the flagship DeepSWE benchmark tests coding agents on multi-file tasks averaging 668 lines of code drawn from 91 open-source repositories across five programming languages, and reporting substantially lower false-positive and false-negative rates than incumbent benchmarks. Supporting products include reinforcement learning environments, off-the-shelf curated datasets, expert agent trajectories, and supervised fine-tuning services spanning software engineering, data science, cybersecurity, and machine learning research domains.
The company operates a community-led, gamified "bounty hunter" platform that pays skilled software engineers to contribute high-quality data, differentiating from traditional crowd-sourced labeling labor pools. Its go-to-market combines direct enterprise sales for custom data contracts with self-service digital access to pre-built datasets and benchmarks, and the company publicly positions against Scale AI in the training-data market. Revenue mechanics span data monetization (curated datasets), subscription/recurring (custom collection and benchmarks), and professional services (supervised fine-tuning), with pricing not publicly disclosed. Founder Serena Ge, who dropped out of the University of Waterloo at age 19, leads the company alongside co-founder Charley Lee.
Datacurve firmographics
Firmographics- Name
- Datacurve
- Legal name
- Datacurve
- Website
- https://datacurve.ai
- Company type
- Private
- Founded year
- 2024
- Operating status
- Operating
- Headcount range
- 11–50 employees
- Short description
- Datacurve is a San Francisco-based AI data infrastructure company founded in 2024 that builds high-quality training datasets, reinforcement learning environments, and evaluation benchmarks, with initial focus on long-horizon software engineering tasks for frontier AI labs.
- Ownership category
- akta.pro rank
Datacurve industry classification
Industry- Product category
- AI Training Data and Evaluation
- NAICS
- Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (518210), Software Publishers (513210)
- SIC
- Services-Computer Processing & Data Preparation (7374), Services-Computer Programming, Data Processing, Etc. (7370)
- akta.pro industry
- Data Labeling & Annotation ServicesHardware, Technology & Telecom/Artificial Intelligence & Machine Learning/Synthetic Data, Data Labeling & Training Data ToolsTraining Data Management, Versioning & LineageHardware, Technology & Telecom/Artificial Intelligence & Machine Learning/Synthetic Data, Data Labeling & Training Data ToolsData Pipelines for GenAI (Curation, Filtering, Deduplication, Copyright)Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/Foundation Models & Generative AI (LLMs, multimodal)
Keywords
Where Datacurve is headquartered
LocationHeadquarters
- HQ city
- San Francisco
- HQ country
- United States
- HQ region
- North America
Offices1 record
Markets served
Datacurve business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Personnel, Technology or R&D, Operations, Infrastructure, Marketing or Sales
Revenue model
- OTS Datasets: Prebuilt datasets curated for signal, reviewed for quality, and structured to drop into training stacks. Revenue from selling access to curated AI training datasets.
- Custom Data Collection: Bespoke data collection services across multiple domains including Software Engineering, Data Science, Cyber Security, Machine Learning Research. Revenue from enterprise contracts for specialized training data.
- Benchmarks and Evals: Quality benchmarks and evaluation services that capture task-faithful, domain-sensitive lift. Sold as evaluation tools for AI models.
- Supervised Fine-tuning: Expert demonstrations captured through bespoke tooling for model training, provided as a service.
Go-to-market motion1 record
Distribution channels2 records
Marketing channels5 records
Datacurve product offering
Product offeringCore offering
Datacurve produces and sells high-quality AI training data, with a primary focus on long-horizon software engineering tasks, using a gamified bounty platform that attracts skilled engineers to contribute expert code, annotations, and trajectories. Its commercial catalog includes OTS prebuilt datasets, custom data collection engagements, supervised fine-tuning demonstrations, agent trajectories, reinforcement learning environments, and benchmarks/evaluations such as the DeepSWE coding benchmark.
Product overview
Datacurve is an AI data infrastructure company operating as a unified platform combining multiple interconnected products and services for AI training data. The core portfolio includes the DeepSWE benchmark for evaluating coding models, reinforcement learning environments for measuring agentic capabilities, long-horizon tasks, prebuilt OTS datasets, agent trajectories, and supervised fine-tuning services. The company targets multiple domains including Software Engineering, Data Science, Cyber Security, and Machine Learning Research. A bounty hunter/gamification system incentivizes skilled software engineers to contribute high-quality data, positioning Datacurve as a competitor to Scale AI in the AI training data market.
Differentiator
Problem solved
Functional benefit
Brands
- DeepSWE: Long-horizon coding benchmark created by Datacurve to evaluate AI coding models using contamination-free tasks drawn from 91 open source repositories across five programming languages.
Products and services
- DeepSWE Benchmark A long-horizon software engineering benchmark that evaluates AI coding models using contamination-free tasks drawn from 91 open source repositories across five programming languages, with rigorous verification yielding 0.3% false positives and 1.1% false negatives; reference solutions average 668 lines of code. Sold to AI labs and frontier model developers seeking trustworthy coding evaluations.
- Reinforcement Learning Environments Durable environments for measuring agentic capabilities, featuring long horizons, natural instructions, realistic tools, domain-specific judgment, and partial progress. Spans enterprise data systems with layered schemas, enterprise software with evolving business logic, and integration-constrained environments across Software Engineering, Data Science, Cyber Security, and Machine Learning Research.
- Long-Horizon Tasks Tasks stretching hours and days that span ambiguity, partial progress, and recovery, collected to reflect real-world work while preserving the signal needed to improve agent behavior. Targeted at AI teams training long-horizon agents.
- OTS Datasets Prebuilt datasets curated for signal, reviewed for quality, and structured to drop directly into training stacks without translation work. Sold to AI teams needing ready-to-use training corpora.
- Benchmarks and Evals Quality benchmarks that capture task-faithful, domain-sensitive lift, designed to track how models actually perform rather than how metric numbers move. Sold to AI labs, model developers, and enterprise AI teams as evaluation tooling.
- Agent Trajectories Full traces of expert execution, including tool calls, checks, pivots, and recoveries, so agents learn the shape of real work and not just the finished answer. Sold to teams training AI agents.
- Supervised Fine-Tuning Expert demonstrations captured through bespoke tooling that lets experts work naturally while preserving the operating judgment behind the work, used to set the right behavioral prior for models. Sold to AI teams needing high-quality fine-tuning data.
- Custom Data Collection Bespoke data collection services across Software Engineering, Data Science, Cyber Security, and Machine Learning Research, with planned expansion into finance, marketing, and medicine. Sold via enterprise contracts to teams needing domain-specific training data.
Quantifiable outcome
- DeepSWE benchmark achieves 0.3% false positives and 1.1% false negatives compared to 8% false positives and 25% false negatives in SWE-Bench Pro
- +1 more outcomes
Companies that use Datacurve
Customer profileNamed customers1 record
Segments4 records
Ideal customer profiles2 records
Datacurve technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- No
- API docs
- API detail
Core technology
AI maturity
App detail
AI capability5 records
Feature3 records
Datacurve partnerships and signals
Strategic signalPartnerships
Two partnerships are on record, tiered minor.
- PintoursminorStrategic partnership combining fan intelligence with AI-powered tour experiences. Enables sports fans attending major events to discover personalized local experiences in host cities. Creates a new category of sports-based travel discovery and monetization channel for host cities, sports organizations, and experience providers. Plans to scale across major sporting moments including World Cup and Olympic Games.
- Dubai Sports CouncilminorPartnership to develop a digital platform for Dubai Open for Football Academies (DOFA). Aims to enhance athlete development, analytics, engagement, and sponsorship through creating verified athlete identities, real-time performance tracking, fan engagement tools, and sponsor activations. Aligns with Dubai's National Sports Strategy 2031.
Scale indicators3 records
Recent moves6 records
Expansion highlights6 records
Datacurve competitors and assessment
Company assessmentDirect peers
- Scale AI: The dominant AI training data platform serving frontier labs with labeled data, RLHF, and evaluation. Datacurve's own Series A announcement explicitly positions the company as a competitor to Scale AI, particularly in high-quality code data.
- Surge AI: Premium data labeling platform focused on high-quality annotation for foundation models, with an emphasis on expert contributors and complex reasoning tasks. Closely comparable to Datacurve's positioning around skilled-talent-driven data quality.
- Snorkel AI: Data-centric AI platform offering programmatic data labeling, training data management, and evaluation tooling for enterprise ML teams. Comparable to Datacurve's data management and enterprise evals portfolio.
- Labelbox: Training data platform combining annotation tools, data management, and model evaluation for AI teams. Overlaps with Datacurve across data labeling, OTS datasets, and benchmarking use cases.
- Toloka: Crowdsourced data labeling and annotation platform serving AI developers with human-in-the-loop data production. Comparable to Datacurve's community-driven bounty approach to data collection.
- Encord: Data labeling, curation, and evaluation platform purpose-built for multimodal AI workflows including computer vision and RLHF. Overlaps with Datacurve's training data management and evaluation offerings.
- SuperAnnotate: End-to-end data annotation and management platform with a marketplace of expert annotators, serving AI teams building production models. Closely comparable to Datacurve's combination of curated datasets and quality-controlled annotation.
- CloudFactory: Managed data labeling and AI data operations provider combining a trained workforce with tooling for computer vision and NLP. Comparable to Datacurve's custom data collection and enterprise data services motion.
Broad incumbents
- Appen: Large, publicly traded data annotation company with a global crowd workforce serving enterprise AI and ML needs. Comparable in offering breadth and customer type but at much larger scale and lower specialization than Datacurve.
- Innodata: Established data preparation and annotation services provider serving large enterprises and AI initiatives across multiple verticals. Comparable to Datacurve's custom data collection services but with a traditional services orientation.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat3 records
Key risks6 records
Key highlights6 records
Customer concentration
Datacurve social profiles
Digital presenceDatacurve financial estimates
Financial estimateRevenue estimate
Valuation estimate
Datacurve leadership team
Management profileNumber of profiles
Profiles2 records
Datacurve funding detail
Funding detailFunding overview
Funding rounds3 records
Investors6 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Datacurve M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Datacurve
What does Datacurve do?
Datacurve produces and sells high-quality AI training data, with a primary focus on long-horizon software engineering tasks, using a gamified bounty platform that attracts skilled engineers to contribute expert code, annotations, and trajectories. Its commercial catalog includes OTS prebuilt datasets, custom data collection engagements, supervised fine-tuning demonstrations, agent trajectories, reinforcement learning environments, and benchmarks/evaluations such as the DeepSWE coding benchmark.
Is Datacurve a public or private company?
Datacurve is a private company. It is classified as venture growth investor backed and is currently operating.
When was Datacurve founded?
Datacurve was founded in 2024. It employs 11 to 50 people.
Where is Datacurve based?
Datacurve is headquartered in San Francisco, United States, in the North America region.
How does Datacurve make money?
Four revenue lines are on record. OTS Datasets are the primary driver. The others are custom Data Collection, benchmarks and Evals and supervised Fine-tuning.
Who are Datacurve's main competitors?
Direct peers on record are Scale AI, Surge AI, Snorkel AI, Labelbox, Toloka, Encord, SuperAnnotate and CloudFactory. Broad incumbents are Appen and Innodata.
Does Datacurve have an API?
No public API is recorded for Datacurve.
What industry is Datacurve in?
Datacurve's product category is AI Training Data and Evaluation. Its primary akta.pro industry code is HDAAALAB, Data Labeling & Annotation Services, with a secondary code of HDAAALAD, Training Data Management, Versioning & Lineage. Its NAICS code is 518210 and its SIC code is 7374.