§ 01 — Practice
From building production distributed systems and cloud-native infrastructure, we’re now driving harness engineering to help enterprises unlock new AI capabilities.
The harness is everything except the model: eval suites, CI gates, drift detectors, compliance artifacts, operational telemetry. We build it. We run it. We keep it current when the model ships a new version. Claude practice live. Sprints, not SOWs. Published OSS at github.com/VISystems.
Every decade, a new substrate becomes the next decade of enterprise infrastructure. VISystems has shipped at every inflection since 2007. The technology changes. The engineering discipline — eval-first, telemetered, audit-trailed, no-SOW — does not.
2007–2015
Built and operated Apache Big Data platforms for F1000 enterprises when the stack was still incubating. Production Hadoop at enterprise scale — ingestion pipelines, Spark analytics, Cassandra persistence — shipped as engineering sprints, not advisory decks.
Anchor projects: enterprise Hadoop clusters, Spark pipelines, Cassandra-at-scale deployments
2016–2023
Shifted substrate to Kubernetes and the CNCF ecosystem. Designed multi-cluster platforms, service-mesh topologies, and GitOps delivery pipelines for regulated financial services and large ISVs. Same discipline: eval-first, telemetered, no-SOW.
Anchor projects: multi-cluster K8s platforms, service-mesh rollouts, GitOps CI/CD
2024 → present
Current era. Harness engineering for enterprise AI. Building and operating the guides and sensors that keep agentic systems reliable — eval suites, CI gates, drift detection, compliance artifacts, operational telemetry. Claude practice live, with Eval Cloud as the operational layer we built ourselves.
Anchor: claude-eval-runner OSS, Eval Cloud private beta
We work with enterprises whose compliance and procurement constraints mean we can’t name them here. The list below is anonymized to the level their legal teams are comfortable with.
▌ Why no logos
“We don’t publish customer logos. Their procurement team prefers it that way.”
§ 04How we operate
A VISystems Sprint is two weeks × 100 SEU — Senior Engineering Units, our internal measure of senior engineer effort. Every capability is scoped in SEU against a published Catalog. There is no work-intake portal. There are no SOWs. There is your PMO tool, our Sprints, and Eval Cloud writing the audit row.
§04.1
Eval suites, context architecture, tools, MCP servers, agents. Built test-first against a published spec, version-controlled in your repos, harness-engineered from day one.
§04.2
Production traces, observability, drift detection, capacity planning, eval-as-CI gates. The day-two work that determines whether agents stay in production a year from now.
§04.3
Migration concierge on every new Claude release. Audit ledger by default. SR 11-7 alignment documentation. The work that survives a model risk officer’s review.
§04.4
Quarterly Reliability Reviews. Capability roadmap co-authoring. Your engineering org gets stronger every quarter, not more dependent on us.
§ 05Tiers — term sheet preview
Every tier ships every product feature — SSO, SCIM, the eval harness, MCP registry, audit ledger, all five PMO adapters, full data export. Tiers differ on Sprint allocation, retention, isolation, and delivery targets — never on whether you can do something. No SSO tax. No SCIM tax. No audit-ledger tax.
| Feature | §05.1 Reliability Core Annual subscription | §05.2 Reliability Pro Production scale | §05.3 Reliability Strategic Dedicated tenancy · SR 11-7-aligned |
|---|---|---|---|
| Sprint allocation | Baseline | Scaled | Enterprise |
| SEU capacity | Standard | Expanded | Full |
| Migration lead time | Standard | Priority | Next business day |
| CS cadence | Quarterly | Quarterly + monthly | + named CSM |
| Trace retention * | 12 mo | 24 mo | 84 months · SR 11-7 |
| DB tenancy * | Shared | Shared | Dedicated Neon project |
| Eval volume * | Standard | Scaled | Enterprise |
| Specific capacity, Sprint allocation, and commercial terms are sized during discovery based on your workload profile. Founding-customer terms available on all tiers signed before Eval Cloud GA. | |||
* Eval Cloud platform features — currently in private beta. Bundled with every subscription at GA.
→ See full term sheet at /services · Top-Up Sprint Packs at /services#topups
§ 06Capability Catalog · published index
The Catalog is the entire customer-facing surface. There is no portal where you submit work requests. You raise a ticket in Jira, Linear, ServiceNow, Azure DevOps, or monday.com — pre-installed during Sprint 0 — pick the Catalog item, scoped in SEU, with a target lead time. Your Authorized Sponsor approves. VISystems delivers. Eval Cloud writes the audit row.
BUILD · ESTABLISH OR MIGRATE EVAL HARNESS
BUILD · CONTEXT ARCHITECTURE, MEMORY, TOKEN BUDGET
BUILD · SCOPED, ISOLATED, EVAL-GATED
BUILD · FROM PRD TO PRODUCTION
RUN · TRACES, SLOs, DRIFT, COST
RUN · GITHUB / GITLAB / BITBUCKET
RUN · CACHE ROI, BATCH, MODEL-TIER
RELY · NEW CLAUDE VERSION → RE-BASELINE
RELY · MODEL RISK MGMT DOCUMENTATION
RELY · TRACE-BACKED ROOT-CAUSE WORK
EVOLVE · CFO-READY OUTCOMES REPORT
EVOLVE · BUILDS NEW ADAPTER ON FRAMEWORK
→ Full Catalog with deliverables, evidence artifacts, escalation paths at /offerings
§ 07 Eval Cloud · the operational layer
▌ Built because we needed it · bundled with every subscription
Every VISystems engagement was already running its own evals, drift detection, and audit ledger. We packaged the discipline into a product, kept building it, and now run our practice on it. Eval Cloud is the operational layer underneath every engagement — runs your eval suites, ingests your traces, watches the Claude release cadence, auto-re-baselines on every new version (VISystems-billed, included at no additional cost), reconciles the Capability Catalog against your PMO ticket events, and emits the audit ledger that doubles as SOC 2 evidence and your own model-risk documentation.
Currently in PRIVATE BETA. Public beta Q3 2026. v1 GA target Q1 2027. Pre-GA subscribers receive preferential founding-customer terms and free historical-telemetry migration at GA cutover.
▌ Waiting list — open
Founding-customer subscribers who join before public beta lock preferential terms and free historical-telemetry migration at GA cutover.
Eval Cloud waiting list
Contact dev@visystems.com to join the private beta.
No marketing emails. We respond within 48 hours with a discovery slot or a sized term sheet, depending on what you ask for.
§ 08 Three differentiators
We bet our brand on three load-bearing differentiators.
▌ Differentiator 01
Every VISystems engagement instruments Claude API growth telemetry from Sprint 1. We publish the metrics — anonymised and aggregated — in the VISystems Impact Report so subscribers can benchmark their own trajectory against the cohort. If your API calls aren't growing, we flag the pattern in QBR before you notice it in billing.
→ VISystems Impact Report · published quarterly to subscribers
▌ Differentiator 02
7 Apache 2.0 repos, all built by the same engineers who deliver your Sprints. There is no separate OSS team — the engineers who deliver Sprints write the OSS. The repos are the proof that the discipline we sell is the discipline we practice.
→ Q1 · claude-eval-runner · live
▌ Differentiator 03
MSA + Subscription + published Catalog. Every deliverable is a named Catalog item with a published scope, a Sprint allocation, and an eval contract. No SOWs. No bespoke scoping. No ambiguity about what you bought.
→ Subscription model · 12 Catalog items · 5 PMO tools
§ 09 What we don’t do · published anti-features
shows up in what it refuses to ship.
We don’t scope engagements per-project. You subscribe, you get a Sprint allocation, and you pull from the published Catalog. Scope changes are a Variation Order — one page, countersigned same week.
Replaced by → MSA + Order Form + Variation Order template
Your subscription includes a fixed Sprint allocation. You don’t get a meter running in the background. You don’t get a surprise invoice because an eval suite ran 400 times instead of 40.
Replaced by → Sprint allocation + Top-Up Sprint Packs
We don’t ask you to log into our portal to submit a ticket. We integrate into your PMO tooling — Jira, Linear, Shortcut, Asana, or GitHub Projects — and work where your engineers already work.
Replaced by → Native PMO integration · 5 tools at v1
§ 10 Who builds your Sprint
VISystems doesn’t have a delivery wing and an engineering wing. The engineers who build our OSS toolchain deliver your Sprints. The team behind Eval Cloud’s audit-ledger module is the team installing it on your tenant. Same engineers, same codebase, same standards.
No bench. No ‘principal architect’ who hasn’t shipped in five years. No blended rate card. No transformation slides. Senior engineers, on subscription, on the Catalog, on the eval ledger.
Whether you’re getting Claude into production or keeping it reliable once it’s there — we’re the firm with engineers in your repos by Sprint 1, in your traces by Sprint 2, and in your audit ledger by Sprint 3.
▌ The shape of a firm shows up in who actually does the work.