Platform overview
The discovery engine for enterprise modernization.
PYD ENTERPRISES compiles legacy COBOL, Java EE, .NET and stored procedures into one symbol graph, then extracts the business rules, data contracts and interfaces your new architecture has to honor.
Every executable path accounted for
- Extracted as rules
- 2,914
- Classified as plumbing
- 11,402
- Flagged for review
- 188
Architecture
Six stages, one symbol graph.
Each stage writes an artifact the next one reads, and every artifact keeps a pointer to the file and line range it came from. Step through one scan of a 1.24 million line lending system to see what each stage produces.
STAGE 01 OF 06
Source ingest
Pulls source from PDS exports, Endevor element libraries and Git mirrors into a volume inside your cluster. Code pages are converted once (IBM-1047 to UTF-8), sequence columns are kept so line numbers match your mainframe listings, and every file is hashed into an immutable scan manifest.
IN
- PDS / Endevor export + manifest
- Git mirrors pinned to a SHA
- DB2 and Oracle DDL
- CICS CSD extract
OUT
- Scan manifest (file, hash, dialect)
- Normalized UTF-8 source volume
14,382
files hashed into the manifest
1{2 "scanId": "loans-2026-11-04",3 "sources": 14382,4 "dialects": {5 "ibm-enterprise-cobol-6.4": 2861,6 "copybook": 4127,7 "jcl-zos": 1904,8 "cics-bms": 212,9 "java-ee-6": 5051,10 "db2-ddl": 22711 },12 "codePage": "IBM-1047 -> UTF-8",13 "sha256": "9c1e4f...b07a"14}
Four pillars
Parse it. Extract it. Specify it. Prove it.
PILLAR 1
Multi-language legacy AST parser
Purpose-built front ends for IBM Enterprise and Micro Focus COBOL, CICS and BMS, JCL, VSAM, IMS DB, DB2 embedded SQL, Java EE, Struts, Spring XML, VB6, WebForms, WCF, PL/SQL, T-SQL and Informix 4GL.
PILLAR 2
Business rules to DMN 1.5
Calculations and policy branches lifted out of procedural glue into decision tables an analyst can sign.
PILLAR 3
OpenAPI 3.1 & schema synthesizer
principalBalance: type: string pattern: '^-?\d{1,11}\.\d{2}$' # LN-PRIN-BAL S9(11)V99 COMP-3
PILLAR 4
Behavioral test generation
- avgBal=1499.99 → 12.00
- avgBal=1500.00 → 0.00
- tier=Z → 30.00 (flagged)
Honest boundaries
What PYD ENTERPRISES does not do.
Black-box transpilation turns bad COBOL into unmaintainable Java. We extract clean specifications and rules so your engineers can build the architecture you actually want.
Transpile COBOL into Java line by line
Extracts the rules and contracts so your engineers design clean services instead of inheriting COBOL-shaped Java.
Write the target application
Hands your build team OpenAPI 3.1 contracts, DMN tables, JSON Schema and fixtures to build and test against.
Ask you to trust unreviewed output
Every rule sits beside its source lines in a sign-off view. Analysts approve, correct or reject before anything is exported as a baseline.
Promise 100% rule coverage
Accounts for every executable path: extracted as a rule, classified as plumbing, or flagged for review. Nothing is silently dropped.
Send code to a cloud model
Runs in your data center or private VPC with egress denied. An optional model you host may name rules; it never decides what code does.
Deterministic first
The engine finds the rule. A model may only name it.
Every field in a PYD ENTERPRISES export has a provenance. source and path come from the control-flow graph. when and then come from dataflow over the resolved storage layout.
If you enable a language model running on your own hardware, it writes the human label, and the export records that it did. It cannot add a row, change a condition or decide that a branch is plumbing. Turn it off and the rule inventory is identical; only the labels go back to generated identifiers.
That separation is why the same source produces the same specification every time, and why an auditor can trace each DMN row to the lines that justify it.
1{2 "id": "DDAFEE-0006",3 "decision": "Maintenance fee",4 "row": 6,5 "when": { "tier": "not(\"B\",\"P\")", "avgBalance": "< 25000.00" },6 "then": { "fee": 30.00 },7 "source": "DDAFEE01.cbl:32-35",8 "path": "2100-BASE-FEE / ELSE / ELSE",9 "classifiedBy": "deterministic",10 "label": "Premier and unknown tiers under 25,000.00",11 "labelBy": "customer-hosted model",12 "review": "flagged: catch-all charges unknown codes"13}
Open standards
Every export opens in tools you already own.
No proprietary viewer, no lock-in. Specifications are files you keep, version and review like code.
OpenAPI3.1
REST interface contracts
exportAsyncAPI2.6
Kafka and event contracts
exportDMN1.5
Business decision tables
exportJSON Schema2020-12
Data contracts and constraints
exportPostman Collection2.1
Executable request suites
exportMermaid / PlantUML
ER and sequence diagrams
export- LOC per hour parse throughput on a 32-vCPU node
- ~1M
- LOC per hour parse throughput on a 32-vCPU node
- PYD ENTERPRISES benchmark figure
- typical discovery window, compressed from about 9 months of manual reading
- 3 wks
- typical discovery window, compressed from about 9 months of manual reading
- PYD ENTERPRISES engagement figure
- per LOC: what manual discovery costs in PYD ENTERPRISES' engagement benchmarks
- $0.40 to $0.75
- per LOC: what manual discovery costs in PYD ENTERPRISES' engagement benchmarks
- PYD ENTERPRISES engagement benchmarks
- of source, AST or rules sent outside your environment
- 0 bytes
- of source, AST or rules sent outside your environment
- Deployment architecture
FAQ
Platform questions.
Is PYD ENTERPRISES a transpiler?
No. PYD ENTERPRISES does not convert COBOL into Java or VB6 into C#. It extracts what the code does (rules, data contracts, interfaces, lineage) as open-standard specifications, so your engineers build services designed for the target platform rather than translated from the old one.
What does deterministic mean here?
Given the same source, compiler options and configuration, PYD ENTERPRISES produces the same AST, symbol graph, rule inventory and exports every time. Classification is driven by the control-flow and dataflow graphs, not by sampling a model. If you enable a language model you host, it only writes names and descriptions for rules the engine already located.
How does a reviewer know an extracted rule is right?
Each rule opens in a side-by-side view with the exact source lines highlighted, the inputs it reads, the outputs it writes and the path conditions that lead to it. Reviewers approve, correct or reject it, and the sign-off is recorded with the rule.
Where does PYD ENTERPRISES run?
In your data center or private VPC on AWS, Azure or GCP, installed from signed Docker images or a Helm chart. Air-gapped installs use offline license keys and signed update bundles. No source, AST or rule leaves your environment.
Which outputs can we import into existing tools?
DMN 1.5 opens in DMN modelers and engines such as Camunda and Drools, OpenAPI 3.1 in API tooling and gateways, Postman 2.1 collections in Postman or Newman, Mermaid and PlantUML in most documentation tools, and JSON Schema 2020-12 anywhere JSON validation is used.
See the pipeline run on your own code.
Run PYD ENTERPRISES on 100,000 lines of COBOL or Java in your private VPC and review OpenAPI specs and DMN tables tomorrow.