2020
Catalog and transformation tool
A single tool for cataloging and transforming enterprise data. Both had always been separate products, never designed to work together.
About us
Dataworkz is an Agent OS for enterprise workflows.
Your own engineering, data and IT teams build the agents, on your data, wherever your security posture requires them to run. More than 200 AI applications built on Dataworkz are in production today.
- Founded
- March 2020
- Deployment
- Managed SaaS, dedicated SaaS, or your own VPC
- Compliance
- SOC 2 Type II and CASA Tier 2
Our journey
How we got here
We started out solving a data problem. Agents came later.
2021
Text to Spark SQL
Core transformation engine shipped and reached market. Text to Spark SQL opened a self-service path into enterprise data for users who did not write code.
2022
Landed enterprise customers
First enterprise customers in production, with a unified view across systems their data had been scattered over.
2023
RAG as a service
Pivoted to RAG as a service. The retrieval layer beneath the product had drawn more customer interest than the product itself.
2024
Agents
Retrieval was not enough.
Working with large enterprises, the use cases worth doing all needed structured and unstructured data in the same answer.
What is the status of my order with the white socks?
The socks are in a product catalog. The order status is in a transactional system. The return policy is in a PDF. Retrieval alone cannot answer it, because answering it is not a lookup. It is a workflow. So we started building agents.
Today
Agent OS
What we built instead
Every enterprise workflow asks for two different things. Deterministic execution: sequencing, validation, human approval, the same behavior every run. Probabilistic intelligence: reading a messy document, interpreting an ambiguous instruction.
LLMs are extraordinary at the second and structurally unsuited to the first. You cannot prompt your way out of a property the model does not have.
So the Agent OS runs on a neuro-symbolic engine. A symbolic backbone, the Dataworkz DSL, executes deterministically and logs every step, calling neural intelligence on demand for the parts that need judgment.
Principles
How we build
Deterministic by construction
The symbolic backbone runs the workflow and the model is called into it, never the other way around. Every step is logged and replayable.
Governed at the chunk
Access control at the retrieved-chunk level, not just the document. An agent sees what the person it acts for is allowed to see.
Model-agnostic
Bring your own LLM, embedding model, vector store and data store. Switch without rework. You are not locked to anyone's roadmap.
Measured per agent
Every agent carries a dollar value, rolled up to a program P&L. AI shows up in the numbers a CFO already reads.
Who we are
Meet the founders
Sachin Smotra
Founder & CEOEnterprise software and data infrastructure leader. Equal parts product manager, strategist, engineer and designer. Launched Couchbase Analytics, which became Couchbase’s fastest growing product line.
Nikhil Smotra
Founder & CTOEnterprise data and AI architect who has built large-scale platforms. Built and led a global data engineering team across the U.S., Poland and India as SVP at iQor; earlier pioneered NoSQL and semantic technologies at Lockheed Martin R&D years before they became mainstream.
In production
What customers are running today
200+AI applications in production across enterprises and startups
9 mo to 1 dayfor regulatory data lineage that used to be built in-house
100+connectors to the systems the business already runs on
Security and deployment
You choose where Dataworkz runs and where your data lives
Three deployment models. Pick the one your security posture requires.
Managed SaaS
We run it. Fastest path to a first agent in production.
Dedicated SaaS
Isolated tenancy, managed by us, separated from every other customer.
Your VPC
Dataworkz runs inside your own cloud environment. Your data stays where it already is.
SOC 2 Type IICASA Tier 2Chunk-level RBAC
Full audit loggingTraceability per agent decision
Encryption at rest and in transit
Get started
Bring us a workflow that has to be right every time
30 minutes with our solutions team. Bring the workflow that costs you the most people-hours, and we will show you what it looks like as an agent.
Backed by Engineering Capital and a group of angel investors.