Two products, built because the same problem kept arriving.
Antvia is a data and intelligence platform. Uloborus is a manufacturing system for Indian factories. They are not two editions of one thing and we are not going to call them a suite. Both are live, both have their own sites, and both exist because rebuilding the same integration work on every engagement stopped being defensible.
The two products, and what each one actually is
Different buyers, different problems and different websites, with the team that builds them as the only shared part.
A production-grade lakehouse, already built, deployed into your own cloud
Bronze, silver and gold layers, governance and orchestration across the middle, 300+ sources supported, and AI trained on the gold layer rather than bolted beside it. It runs in your VPC on AWS, Azure, GCP or on-prem.
Read the Antvia page →UloborusManufacturing systemBuying, receiving, quality, stock, making, selling and dispatch on one thread
Eight modules over one set of records, with the GST paperwork travelling with the goods rather than being retyped into a separate accounts package. Built with a working factory rather than for a demo.
Read the Uloborus page →Why a consultancy ends up building products
It happens the same way each time, and it is worth saying plainly rather than dressing it up as a strategy.
- Pattern
The same problem kept arriving
Different industries, different logos, the same shape underneath. Sources that will not join, a quality step nobody enforces, a number Finance cannot trace, a plant running on spreadsheets that quietly became the system of record.
- Repetition
The integration work was rebuilt every engagement
Ingestion, quality checks, reconciliation, lineage capture, access control. Necessary every time, interesting almost never, and paid for out of a budget the client believed was going towards their actual problem.
- Limit
Rebuilding it each time stopped being defensible
If we already know what the first two months look like, charging a client to watch us do it again is hard to justify. Building it once and bringing it with us is the honest version.
- Product
Consulting alone could not carry the load
A pod of three can only be in one place. A platform that is already built travels to the next engagement, which is the only way the six-week median holds. The products are what the consulting learned.
What goes into Antvia, and what reads out of it
Five source types in, five consumer surfaces out, with bronze, silver and gold in between.
- DatabasesDashboards
- App logsAI and ML
- SpreadsheetsReports
- Event streamsMobile
- APIsEmbedded
Antvia's own claims, which you should check
These are published on antvia.io, and since we build and run Antvia we are not a neutral party.
It runs in your cloud, not ours
Deployed into your own VPC on AWS, Azure, GCP or on-prem. Open source under the hood, with open formats, open SQL, open streams and open transforms. Export anytime.
You own the platform rather than renting a seat in one
The distinction Antvia draws with the proprietary vendors is that you are not a tenant inside their engine, with your data held in a format only that engine can read.
Governance is enforced on the query, not written in a policy document
Column-level access control and PII masking are applied on every query, with a governed upload workflow for business users who would otherwise email a spreadsheet.
AI trained on the gold layer, in the box
A decision engine, an inference and predictions API, and natural-language ask, all reading the same modelled layer the dashboards read. Decisions, predictions and chat, without a separate ML team.
Time to first dashboards on your own data
Antvia's published comparison between building a lakehouse yourself and starting from one that already exists.
One thread through the factory
Eight modules over one set of records: buying, receiving, quality, stock, making, selling, dispatch and invoicing.
- Buy
The purchase order is raised
The record that asked for the material, and the one everything downstream still points back to.
- Receive
Goods arrive and quality decides
The receipt lands against the order it belongs to, and the inspection decides whether the material moves or waits.
- Make
A work order runs against a formula
The batch consumes the material it was given and becomes finished goods that still know where they came from.
- Dispatch
The documents leave with the goods
Delivery challan, e-way bill and certificate of analysis, generated from the same records the plant runs on.
- Invoice
The GST invoice comes off those records
Not retyped into a separate accounts package, and not a challan that can be invoiced twice.
Uloborus, in the figures stated on this page
The numbers from the paragraphs above, pulled out where they can be read at a glance.
What holds Uloborus together
Not eight products that sync overnight, but one set of records seen from different desks.
Who it is built for
Process manufacturers of roughly twenty to three hundred people that have outgrown the spreadsheet and cannot stop for six months of implementation. Cosmeceuticals, food, speciality chemicals, pharma ancillaries and contract manufacturers of any of those.
Built with a working factory
Every module was built against how Rajaram Consumer Care, a contract manufacturer of cosmeceutical topicals in Islampur MIDC, Sangli, actually works, then argued about, then changed.
One thread, end to end
A batch traces back through the work order that made it, the material it consumed, the receipt that brought it in and the purchase order that asked for it. A link you follow, in either direction, in seconds.
The paperwork is the product
A delivery challan, an e-way bill, a GST invoice, a certificate of analysis. These are what leave the factory, and they come from the same records the plant runs on.
Every row says who and when
Not an audit module you switch on. The table cannot be written without an author and a timestamp, and every document that changes state keeps the before and after.
It refuses the impossible
You cannot dispatch against an unconfirmed order, consume a batch quality is still holding, or invoice a challan twice. The rules are in the database, not in a training session somebody missed.
A read-only companion
A Flutter app with six screens: today, alerts, stock, orders, money and parties. Read-only is enforced in code, not merely intended. The Android build is done and verified. iOS is configured but not yet compiled.
What the phone deliberately cannot do
The API client hard-codes GET and a test asserts it stays that way. Master data setup, analytics, bins, tax rates, people administration and every write are absent, because a four-inch screen is the worst place to raise a document.
Each step, and the record it leaves behind
The same thread read as paperwork, from the purchase order that asked for the material to the GST invoice that closes it.
- BuyPurchase order
- ReceiveGoods receipt
- MakeWork order and batch
- DispatchDelivery challanE-way billCertificate of analysis
- InvoiceGST invoice
How the two relate, and how they do not
They share a team and a set of engineering habits, and not much else.
They are not a suite
Antvia is bought by a company with data in more places than it can join. Uloborus is bought by a factory that has outgrown the spreadsheet. Different buyers, different sales conversations, separate brands and separate sites.
Uloborus is amber, Woodfrog is blue
That is not an accident of design. If somebody offers you both in one contract because they go well together, be suspicious, including when that somebody is us.
The honest connection is the dataset
Because the rules sit in the database and every row carries its author and timestamp, a plant running Uloborus ends up with a clean, governed operational dataset rather than a decade of reconstructed history.
A direction, not a shipped integration
That dataset is the thing agents could later be built on, and Antvia is where that work would happen. It should be read as intent and nothing more.
What Antvia Intelligence does
Two limits in mainstream BI look structural rather than fixable, and the BI half is built around these five.
Do the interactive work where the user already is
When the whole workload sits on the backend, every filter and drill is a round trip, and cost becomes a function of how curious your team is.
Send to the backend only what genuinely needs it
The rest stays local, so a question does not have to be worth paying for before somebody asks it.
Treat the model as part of the analysis
AI arrived after mainstream BI was designed, so it sits in a panel that cannot see what you are looking at or carry the definitions Finance agreed.
Keep definitions in one place and make them binding
One agreed definition applied everywhere, rather than the same metric quietly drifting between dashboards.
Show the working
Any number traces back to the query and the source behind it.
Where these are the wrong answer
The fastest way to waste a month is to start an implementation that was never going to fit.
Uloborus is wrong for discrete assembly
Deep multi-level bills of material and routings are not what this is built around. It is built around batches and formulas, and stretching it to fit assembly work would be a rebuild, not a configuration.
Uloborus is wrong for a plant with no quality step
Half of what makes the change worth making is holding material until somebody decides. Without that decision point, most of the value is theoretical.
Uloborus is not your books
It produces GST invoices that reconcile, credit notes that reference the original and payments matched against them. It does not do payroll, fixed assets or full financial accounting, and it is not trying to.
Uloborus does not ship AI agents today
Every list view is analytical from day one, with a stats strip, boolean query search and a breakdown panel. Anomaly and margin-leak detection are on the roadmap. No assistant, no copilot, no forecasting today.
Antvia is wrong when your platform is already sound
We work in your tool when your tool is fine. Most of our analytics delivery has happened inside the platform the client already runs, including Apache Superset. You hear that in week one, not month four.
Antvia Intelligence has nothing published you could verify
No benchmarks, no pricing, no release dates and no customer list, because there is nothing yet a Finance team could check. Those get published when a customer's Finance team has checked them, and not before.
The six weeks in which either product goes in
Week 1 is the fixed-fee audit, weeks 2 to 6 are the build, and something is working by week 3.
How we got here
Three years in, from Pune, delivering across APAC, India and the US.
- 2023
Founded in Pune
And the first enterprise client onboarded the same year.
- 2024
Expanded into AI agent development
The agent work is where the gap between a clean dataset and a usable one became impossible to ignore.
- 2025
Serving clients across industries
Fintech, B2B SaaS, retail and public sector, with the same integration groundwork repeating underneath most of it.
- Now
Two products live
Antvia and Uloborus, built and run by the same engineers who do the consulting. Woodfrog is an Anthropic Build Partner shipping production AI agents on Claude, and joined the Claude Partner Network at launch as a Registered partner.
Where to read the detail
Our pages describe how each product fits an engagement, and the product sites are the products' own words.
The platform, the BI half, and how it enters an engagement
Architecture, governance, the published cost and time figures, and where it is the wrong answer.
Antvia on Woodfrog →Product siteantvia.ioAntvia's own site, including the animated tour
Five sources flowing in, bronze promoted to silver, silver modelled into gold, dashboards and chat reading live, and the benchmark report on request from hello@antvia.io.
Visit antvia.io →Woodfrog pageUloborusThe eight modules, the mobile companion, and who it suits
How the thread holds from purchase order to invoice, and the three cases where it is the wrong system.
Uloborus on Woodfrog →Product siteuloborus.comUloborus's own site and signup
The intelligent web of manufacturing. No card, and your own workspace in about a minute.
Visit uloborus.com →The Antvia numbers, as Antvia publishes them
Published on antvia.io rather than measured by us for this page.
Questions buyers ask about the products
Are Antvia and Uloborus the same product with different labels?
No. Antvia is a data lakehouse and intelligence platform deployed into your cloud. Uloborus is an operations system for a factory floor, with GST documents coming out of it. Nothing about buying one makes the other cheaper or easier.
Do we have to take the consulting to use either product?
Antvia is deployed and run by Woodfrog engineers, so in practice that is an engagement. Uloborus you can sign up for on its own site, with no card and a workspace in about a minute.
Who owns the data, and what happens if we leave?
Antvia is deployed in your own VPC on AWS, Azure, GCP or on-prem, open source under the hood, with open formats, open SQL, open streams and open transforms. Export anytime.
Does Uloborus do AI?
Not in the way most vendors mean it. Every list view is analytical from day one, with a stats strip, boolean query search and a breakdown panel. Anomaly and margin-leak detection are on the roadmap. No assistant or forecasting today.
Can we see the Antvia benchmark before we commit?
Yes. It covers 80+ benchmarking metrics across every layer of the stack, measured on 480 GB across 3.2 billion rows and twelve source systems, head to head against AWS Athena, Snowflake, Google BigQuery and Databricks. Ask hello@antvia.io. No form, no tracking, and it includes methodology, sample SQL and an optional walkthrough.
We already have Snowflake, Databricks or Fabric. Why would we move?
Often you would not, and that answer is free. The case Antvia makes is about the shape of the bill: compute billed per query or per hour, seats with contract minimums, governance and AI behind premium tiers, egress fees, multi-year spend floors.
Who actually builds and supports these?
A pod of three experienced practitioners, no bench, nobody learning on your clock. Week 1 is the fixed-fee audit, weeks 2 to 6 are build with something working by week 3, and week 7 onward is quarterly reviews and on-call governance.
Start with the audit, not with a product demo
Two calls, a fixed fee, and a one-pager naming what is broken, what it is costing you and the order to fix it in. You keep it either way, including when the answer is that neither product fits. Write to hello@woodfrog.tech.