Ship AI-nativedata experiences.
Kaizen AI Systems is a focused stack of three products: Akuma (chat with your database), Enzan (real-time GPU cost clarity), and Sōzō (synthetic data you can trust).
- Products
- Akuma · Enzan · Sōzō
- Teams
- Data · Infra · AI/ML
- SDKs
- TS · Python · Go
// One client, three products
import { akuma, enzan, sozo } from "@kaizen/sdk";
// Akuma: chat with your database
const query = await akuma.query({
prompt: "Top 10 customers by MRR",
dialect: "postgres"
});
// Enzan: real-time GPU cost
const spend = await enzan.summary({ window: "24h" });
// Sōzō: synthetic data
const data = await sozo.generate({
records: 50000,
schema: "saas_customers_v1"
});The Kaizen Stack
Three sharp tools. One mental model.
Akuma
Chat with your database, not with SQL.
Natural language → safe, optimized SQL over your warehouses and OLTP databases.
Enzan
Real-time clarity for your GPU spend.
Unify GPU usage and cost across clouds and vendors in one place.
Sōzō
Synthetic data you can trust.
Generate realistic, privacy-safe tabular datasets for testing and ML.
Chat with your database,
not with SQL.
Akuma ingests your schema and generates SQL that respects your dialect, naming conventions, and guardrails.
- Schema-aware NL→SQL for Postgres, MySQL, Snowflake, BigQuery
- Read-only enforcement and table/column allowlists
- Modes: SQL-only, SQL+results, or explain query
- Easy to embed into dashboards or chat UIs
import { akuma } from "@kaizen/sdk";
const { sql, rows } = await akuma.query({
dialect: "postgres",
prompt: "Show MRR by plan for last 6 months",
mode: "sql-and-results",
guardrails: {
readOnly: true,
allowTables: ["invoices", "plans"],
},
});
console.log(sql);
// SELECT date_trunc('month', invoice_date)...import { enzan } from "@kaizen/sdk";
const summary = await enzan.summary({
window: "24h",
groupBy: ["project", "model"],
});
console.table(summary.rows);
// [
// { project: "fraud-api", cost_usd: 182.33 },
// { project: "support-bot", cost_usd: 97.54 },
// ]Real-time clarity for
your GPU spend.
Enzan pulls metrics from your GPU providers and turns them into clean, queryable telemetry.
- Normalize usage and cost across clouds
- Attribute spend by project, model, or team
- Build dashboards or export to warehouse
- Great for AI infra and finance teams
Production-like data
without the risk.
Sōzō generates realistic, statistically grounded synthetic data without copying real user records.
- Schema-based generation for tabular data
- Control distributions and correlations
- Export to Parquet/CSV/JSONL
- Fits into CI pipelines and staging
from kaizen import sozo
data = sozo.generate(
records=50000,
schema={
"user_id": "uuid4",
"plan": "choice:free,pro,enterprise",
"mrr": "float:0-500",
"churned": "boolean:0.18",
},
correlations={"plan:mrr": "positive"},
)
data.to_parquet("customers.parquet")Pricing
One platform key. Predictable pricing.
Start free on all three products, then pay only for what you use.
Builder
For solo builders and small teams.
No credit card required
- 1k Akuma queries / month
- 24h Enzan lookback, 5 projects
- 10k Sōzō records / month
- Community support
Production
For real traffic, teams, and SLAs.
Volume discounts available
- Volume discounts per product
- Dedicated rate limits & regions
- Audit logs, SSO, SOC2-ready
- Priority support & onboarding
Docs
Curl, then wire it into your stack.
# Akuma – NL → SQL
curl https://api.kaizenaisystems.com/v1/akuma/query \
-H "Authorization: Bearer $KAIZEN_API_KEY" \
-d '{"dialect":"postgres","prompt":"Top 10 by MRR"}'
# Enzan – GPU spend
curl https://api.kaizenaisystems.com/v1/enzan/summary \
-H "Authorization: Bearer $KAIZEN_API_KEY" \
-d '{"window":"24h","groupBy":["project"]}'
# Sōzō – Synthetic data
curl https://api.kaizenaisystems.com/v1/sozo/generate \
-H "Authorization: Bearer $KAIZEN_API_KEY" \
-d '{"schemaName":"saas_customers_v1","records":1000}'- Read-only helpers and guardrails for Akuma
- Cost ceilings and alerts for Enzan
- Schema and privacy controls for Sōzō