| ---
|
| license: mit
|
| language:
|
| - en
|
| pretty_name: Semanta Dataset Suite - Industry Vertical Synthetic Worlds
|
| task_categories:
|
| - tabular-classification
|
| - tabular-regression
|
| - time-series-forecasting
|
| tags:
|
| - semanta
|
| - synthetic-data
|
| - world-native-data
|
| - tabular
|
| - banks
|
| - ai
|
| - hedge_funds
|
| - insurance
|
| - pharma
|
| - medicine
|
| - manufacturing
|
| - security
|
| - science
|
| - fmcg
|
| - retail
|
| size_categories:
|
| - 1M<n<10M
|
| ---
|
|
|
| # Semanta Dataset Suite
|
|
|
| Semanta is a **World Intelligence Operating System**. This package is a market-facing proof that Semanta can generate
|
| large, structured, scenario-covered synthetic worlds for many industries, not just one narrow demo dataset.
|
|
|
| ## Start Here
|
|
|
| | Need | Open |
|
| |---|---|
|
| | See the public product surface | https://semanta.xyz |
|
| | Run the no-login proof bundle | https://api-staging.semanta.xyz/public/proof-bundle?rows=64&seed=37 |
|
| | Open the operator-style Guided Demo | https://studio-staging.semanta.xyz |
|
| | Inspect the industry-suite API proof | https://api-staging.semanta.xyz/public/industry-dataset-suite |
|
| | Read API docs | https://api-staging.semanta.xyz/docs |
|
| | Download the enterprise brief | docs/Semanta_Enterprise_Brief.pdf |
|
| | Download the investor overview | docs/Semanta_Investor_Overview.pdf |
|
|
|
| ## Customer / Investor Proof Path
|
|
|
| | Step | Proof |
|
| |---:|---|
|
| | 1 | Open `semanta.xyz` and check the public `13/13` proof gate. |
|
| | 2 | Open Studio and run `Guided Demo` for the live public-safe workflow. |
|
| | 3 | Inspect this HF branch for the published dataset suite and quality metrics. |
|
| | 4 | Open the proof bundle for machine-readable evidence, lineage and claim boundaries. |
|
|
|
| ## Executive Snapshot
|
|
|
| | Metric | Value |
|
| |---|---:|
|
| | Run ID | `industry-suite-20260618-01` |
|
| | Canon | `v3.7 Final` |
|
| | Total published synthetic rows in repo | 2,370,621 |
|
| | Industry vertical suite rows | 1,100,000 |
|
| | Industries | 11 |
|
| | Rows | 1,100,000 |
|
| | Columns per industry | 100 |
|
| | Total cells | 110,000,000 |
|
| | Overall quality score | 0.965 |
|
| | Scenario coverage score | 1.000 |
|
| | Production smoke gate | 13/13 public proof targets passed |
|
| | Synthetic-only | yes |
|
| | External provider used | no |
|
|
|
| ## What This Proves
|
|
|
| - Semanta can produce broad, structured, HF-compatible synthetic datasets, not only narrow demo files.
|
| - Every industry pack ships with data, schema, quality metrics, lineage and reproducibility evidence.
|
| - The public proof path currently passes 13/13 production smoke targets, including API health, public demo, proof bundle, Studio Guided Demo contracts, landing and HF README.
|
| - The suite is synthetic-only: no customer data is used, and no customer data is sent to DeepSeek, Gamma or any external model provider.
|
| - The data is designed for ML/DL/NN/Q experiments, scenario robustness, drift prototypes and StarForge/Gamma training handoff.
|
|
|
| ## What You Can Download
|
|
|
| | Artifact | Why it matters |
|
| |---|---|
|
| | `data/industry_vertical_suite/<industry>/*.csv.gz` | Training-ready synthetic tabular worlds by industry. |
|
| | `schemas/industry_vertical_suite/*.json` | Machine-readable column contracts and semantic types. |
|
| | `metrics/industry_vertical_suite/*_quality_metrics.json` | Per-vertical statistical quality and coverage metrics. |
|
| | `metrics/industry_vertical_suite/suite_scorecard.json` | Executive suite-level readiness and quality scorecard. |
|
| | `semanta/industry_vertical_suite/lineage.json` | Reproducibility and source-policy evidence. |
|
| | `semanta/industry_vertical_suite/publication_manifest.json` | Publication manifest, files and claim boundaries. |
|
| | `docs/Semanta_Enterprise_Brief.pdf` | Lightweight enterprise/investor PDF explaining Semanta capabilities. |
|
| | `docs/Semanta_Investor_Overview.pdf` | Short investor-facing Semanta overview. |
|
|
|
| ## Market Positioning
|
|
|
| A company's real data is often one observed degree out of a much larger 360-degree operating space. Semanta expands that narrow slice into scenario-covered, risk-aware worlds so ML/DL/NN/Q systems can train and test against more of the possible reality before reality charges the full cost of failure.
|
|
|
| The practical idea: real company data is often only a thin observed slice of possible reality. Semanta expands that
|
| slice into controlled scenarios, shocks, regimes and tail events. In geometric terms, the observed dataset may be one
|
| degree on one plane; the operating world contains many planes, regimes and trajectories. Semanta makes that hidden
|
| space explorable and trainable.
|
|
|
| ## Industry Packs
|
|
|
| | Industry | Slug | Rows | Columns | Quality | Primary use case |
|
| |---|---:|---:|---:|---:|---|
|
| | Banks | `banks` | 100,000 | 100 | 0.968 | credit risk, transaction anomaly, liquidity and operational-risk simulation |
|
| | AI | `ai` | 100,000 | 100 | 0.973 | prompt complexity, eval difficulty, hallucination risk and tool-use scenario generation |
|
| | Hedge Funds | `hedge_funds` | 100,000 | 100 | 0.978 | factor regime, drawdown, liquidity and alpha-decay scenario generation |
|
| | Insurance | `insurance` | 100,000 | 100 | 0.977 | claim frequency, loss severity, reserve adequacy and catastrophe-tail simulation |
|
| | Pharma | `pharma` | 100,000 | 100 | 0.961 | trial signal, safety, regulatory-delay and launch-readiness scenario generation |
|
| | Medicine | `medicine` | 100,000 | 100 | 0.967 | patient-flow, diagnostic uncertainty, treatment response and capacity-risk simulation |
|
| | Manufacturing | `manufacturing` | 100,000 | 100 | 0.957 | defect, downtime, sensor drift, throughput and supplier-risk simulation |
|
| | Security | `security` | 100,000 | 100 | 0.981 | attack intensity, alert noise, response-latency and asset-criticality simulation |
|
| | Science | `science` | 100,000 | 100 | 0.946 | hypothesis strength, experiment noise, replication and discovery-potential simulation |
|
| | FMCG | `fmcg` | 100,000 | 100 | 0.952 | promotion, demand elasticity, stockout, channel-shift and pricing scenario generation |
|
| | Retail | `retail` | 100,000 | 100 | 0.957 | basket, churn, traffic, markdown, inventory and customer-behavior simulation |
|
|
|
| ## Industry Business Catalog
|
|
|
| | Industry | Buyer | Business question Semanta helps answer | Output artifacts |
|
| |---|---|---|---|
|
| | Banks | banks, fintech risk teams, credit and fraud operations | What happens to credit, fraud, liquidity and operational-risk models when the regime changes before historical evidence is abundant? | `data/industry_vertical_suite/banks/semanta_banks_world_native_synthetic_100000x100.csv.gz` + metrics + schema |
|
| | AI | AI labs, model builders, evaluation teams, agent-platform teams and enterprise AI groups | Can models and agents be evaluated on controlled task worlds before real customer traffic and private data are involved? | `data/industry_vertical_suite/ai/semanta_ai_world_native_synthetic_100000x100.csv.gz` + metrics + schema |
|
| | Hedge Funds | hedge funds, quant teams, portfolio researchers and systematic trading groups | Where do strategies break under liquidity shocks, crowding, drawdowns and factor regime rotation? | `data/industry_vertical_suite/hedge_funds/semanta_hedge_funds_world_native_synthetic_100000x100.csv.gz` + metrics + schema |
|
| | Insurance | insurers, reinsurers, actuarial teams and claim operations | How do underwriting, reserve and claims models behave under rare loss clusters and catastrophe-like tails? | `data/industry_vertical_suite/insurance/semanta_insurance_world_native_synthetic_100000x100.csv.gz` + metrics + schema |
|
| | Pharma | pharmaceutical R&D, clinical operations, safety and market-access teams | Which trial, safety, regulatory-delay and launch-readiness signals deserve attention before clinical capacity is committed? | `data/industry_vertical_suite/pharma/semanta_pharma_world_native_synthetic_100000x100.csv.gz` + metrics + schema |
|
| | Medicine | health systems, hospitals, medtech teams and clinical AI builders | Can triage, readmission and capacity models be tested under pressure without exposing patient data? | `data/industry_vertical_suite/medicine/semanta_medicine_world_native_synthetic_100000x100.csv.gz` + metrics + schema |
|
| | Manufacturing | industrial operators, process engineers, quality and supply-chain teams | Which sensor drift, defect, downtime and supplier-risk pathways should be rehearsed before real production loss? | `data/industry_vertical_suite/manufacturing/semanta_manufacturing_world_native_synthetic_100000x100.csv.gz` + metrics + schema |
|
| | Security | cybersecurity vendors, SOC teams, fraud defense and critical-infrastructure defenders | Can detection and response pipelines withstand adversarial incidents, alert floods and analyst overload? | `data/industry_vertical_suite/security/semanta_security_world_native_synthetic_100000x100.csv.gz` + metrics + schema |
|
| | Science | research labs, frontier R&D groups, scientific ML teams and grant-backed institutions | Which hypotheses, experiments and replication risks should be prioritized before scarce lab time is spent? | `data/industry_vertical_suite/science/semanta_science_world_native_synthetic_100000x100.csv.gz` + metrics + schema |
|
| | FMCG | consumer goods manufacturers, category teams, pricing and demand planners | How will promotion, demand elasticity, stockout and channel-shift decisions behave before trade-spend is committed? | `data/industry_vertical_suite/fmcg/semanta_fmcg_world_native_synthetic_100000x100.csv.gz` + metrics + schema |
|
| | Retail | retailers, marketplaces, CRM, inventory and merchandising teams | What happens to churn, baskets, markdowns, store traffic and inventory when customer behavior drifts? | `data/industry_vertical_suite/retail/semanta_retail_world_native_synthetic_100000x100.csv.gz` + metrics + schema |
|
|
|
| ## Industry Example Cards
|
|
|
| ### 1. Banks
|
|
|
| - **Buyer:** banks, fintech risk teams, credit and fraud operations.
|
| - **Business question:** What happens to credit, fraud, liquidity and operational-risk models when the regime changes before historical evidence is abundant?
|
| - **World modeled:** credit risk, transaction anomaly, liquidity and operational-risk simulation.
|
| - **Success story:** A bank can stress-test credit and fraud models against synthetic regime shifts before the same pattern appears in production.
|
| - **Download:** `data/industry_vertical_suite/banks/semanta_banks_world_native_synthetic_100000x100.csv.gz`.
|
| - **Evidence:** `metrics/industry_vertical_suite/banks_quality_metrics.json` and `schemas/industry_vertical_suite/banks_schema.json`.
|
| - **Shape:** 100,000 rows x 100 columns; quality score 0.968.
|
|
|
| ### 2. AI
|
|
|
| - **Buyer:** AI labs, model builders, evaluation teams, agent-platform teams and enterprise AI groups.
|
| - **Business question:** Can models and agents be evaluated on controlled task worlds before real customer traffic and private data are involved?
|
| - **World modeled:** prompt complexity, eval difficulty, hallucination risk and tool-use scenario generation.
|
| - **Success story:** An AI team can generate controlled evaluation/training substrates for agents and models while keeping customer data out of external model providers.
|
| - **Download:** `data/industry_vertical_suite/ai/semanta_ai_world_native_synthetic_100000x100.csv.gz`.
|
| - **Evidence:** `metrics/industry_vertical_suite/ai_quality_metrics.json` and `schemas/industry_vertical_suite/ai_schema.json`.
|
| - **Shape:** 100,000 rows x 100 columns; quality score 0.973.
|
|
|
| ### 3. Hedge Funds
|
|
|
| - **Buyer:** hedge funds, quant teams, portfolio researchers and systematic trading groups.
|
| - **Business question:** Where do strategies break under liquidity shocks, crowding, drawdowns and factor regime rotation?
|
| - **World modeled:** factor regime, drawdown, liquidity and alpha-decay scenario generation.
|
| - **Success story:** A quant team can evaluate strategy fragility across liquidity shocks and factor crowding without waiting for the next real drawdown.
|
| - **Download:** `data/industry_vertical_suite/hedge_funds/semanta_hedge_funds_world_native_synthetic_100000x100.csv.gz`.
|
| - **Evidence:** `metrics/industry_vertical_suite/hedge_funds_quality_metrics.json` and `schemas/industry_vertical_suite/hedge_funds_schema.json`.
|
| - **Shape:** 100,000 rows x 100 columns; quality score 0.978.
|
|
|
| ### 4. Insurance
|
|
|
| - **Buyer:** insurers, reinsurers, actuarial teams and claim operations.
|
| - **Business question:** How do underwriting, reserve and claims models behave under rare loss clusters and catastrophe-like tails?
|
| - **World modeled:** claim frequency, loss severity, reserve adequacy and catastrophe-tail simulation.
|
| - **Success story:** An insurer can rehearse reserve and claim workflows against rare loss clusters before portfolio stress becomes visible in historical data.
|
| - **Download:** `data/industry_vertical_suite/insurance/semanta_insurance_world_native_synthetic_100000x100.csv.gz`.
|
| - **Evidence:** `metrics/industry_vertical_suite/insurance_quality_metrics.json` and `schemas/industry_vertical_suite/insurance_schema.json`.
|
| - **Shape:** 100,000 rows x 100 columns; quality score 0.977.
|
|
|
| ### 5. Pharma
|
|
|
| - **Buyer:** pharmaceutical R&D, clinical operations, safety and market-access teams.
|
| - **Business question:** Which trial, safety, regulatory-delay and launch-readiness signals deserve attention before clinical capacity is committed?
|
| - **World modeled:** trial signal, safety, regulatory-delay and launch-readiness scenario generation.
|
| - **Success story:** A pharma team can simulate trial-operation bottlenecks and adverse-event signal drift before allocating expensive clinical capacity.
|
| - **Download:** `data/industry_vertical_suite/pharma/semanta_pharma_world_native_synthetic_100000x100.csv.gz`.
|
| - **Evidence:** `metrics/industry_vertical_suite/pharma_quality_metrics.json` and `schemas/industry_vertical_suite/pharma_schema.json`.
|
| - **Shape:** 100,000 rows x 100 columns; quality score 0.961.
|
|
|
| ### 6. Medicine
|
|
|
| - **Buyer:** health systems, hospitals, medtech teams and clinical AI builders.
|
| - **Business question:** Can triage, readmission and capacity models be tested under pressure without exposing patient data?
|
| - **World modeled:** patient-flow, diagnostic uncertainty, treatment response and capacity-risk simulation.
|
| - **Success story:** A hospital can evaluate triage and readmission models under capacity shock without exposing patient data to external systems.
|
| - **Download:** `data/industry_vertical_suite/medicine/semanta_medicine_world_native_synthetic_100000x100.csv.gz`.
|
| - **Evidence:** `metrics/industry_vertical_suite/medicine_quality_metrics.json` and `schemas/industry_vertical_suite/medicine_schema.json`.
|
| - **Shape:** 100,000 rows x 100 columns; quality score 0.967.
|
|
|
| ### 7. Manufacturing
|
|
|
| - **Buyer:** industrial operators, process engineers, quality and supply-chain teams.
|
| - **Business question:** Which sensor drift, defect, downtime and supplier-risk pathways should be rehearsed before real production loss?
|
| - **World modeled:** defect, downtime, sensor drift, throughput and supplier-risk simulation.
|
| - **Success story:** A plant can train predictive quality and downtime models on rare failure pathways without breaking real production equipment.
|
| - **Download:** `data/industry_vertical_suite/manufacturing/semanta_manufacturing_world_native_synthetic_100000x100.csv.gz`.
|
| - **Evidence:** `metrics/industry_vertical_suite/manufacturing_quality_metrics.json` and `schemas/industry_vertical_suite/manufacturing_schema.json`.
|
| - **Shape:** 100,000 rows x 100 columns; quality score 0.957.
|
|
|
| ### 8. Security
|
|
|
| - **Buyer:** cybersecurity vendors, SOC teams, fraud defense and critical-infrastructure defenders.
|
| - **Business question:** Can detection and response pipelines withstand adversarial incidents, alert floods and analyst overload?
|
| - **World modeled:** attack intensity, alert noise, response-latency and asset-criticality simulation.
|
| - **Success story:** A security team can harden detection pipelines against adversarial behavior and alert floods before analysts are overloaded in production.
|
| - **Download:** `data/industry_vertical_suite/security/semanta_security_world_native_synthetic_100000x100.csv.gz`.
|
| - **Evidence:** `metrics/industry_vertical_suite/security_quality_metrics.json` and `schemas/industry_vertical_suite/security_schema.json`.
|
| - **Shape:** 100,000 rows x 100 columns; quality score 0.981.
|
|
|
| ### 9. Science
|
|
|
| - **Buyer:** research labs, frontier R&D groups, scientific ML teams and grant-backed institutions.
|
| - **Business question:** Which hypotheses, experiments and replication risks should be prioritized before scarce lab time is spent?
|
| - **World modeled:** hypothesis strength, experiment noise, replication and discovery-potential simulation.
|
| - **Success story:** A lab can prioritize experiments by testing hypothesis robustness and replication risk before spending scarce lab time.
|
| - **Download:** `data/industry_vertical_suite/science/semanta_science_world_native_synthetic_100000x100.csv.gz`.
|
| - **Evidence:** `metrics/industry_vertical_suite/science_quality_metrics.json` and `schemas/industry_vertical_suite/science_schema.json`.
|
| - **Shape:** 100,000 rows x 100 columns; quality score 0.946.
|
|
|
| ### 10. FMCG
|
|
|
| - **Buyer:** consumer goods manufacturers, category teams, pricing and demand planners.
|
| - **Business question:** How will promotion, demand elasticity, stockout and channel-shift decisions behave before trade-spend is committed?
|
| - **World modeled:** promotion, demand elasticity, stockout, channel-shift and pricing scenario generation.
|
| - **Success story:** An FMCG team can rehearse promotion and channel-mix decisions across demand shocks before committing trade-spend.
|
| - **Download:** `data/industry_vertical_suite/fmcg/semanta_fmcg_world_native_synthetic_100000x100.csv.gz`.
|
| - **Evidence:** `metrics/industry_vertical_suite/fmcg_quality_metrics.json` and `schemas/industry_vertical_suite/fmcg_schema.json`.
|
| - **Shape:** 100,000 rows x 100 columns; quality score 0.952.
|
|
|
| ### 11. Retail
|
|
|
| - **Buyer:** retailers, marketplaces, CRM, inventory and merchandising teams.
|
| - **Business question:** What happens to churn, baskets, markdowns, store traffic and inventory when customer behavior drifts?
|
| - **World modeled:** basket, churn, traffic, markdown, inventory and customer-behavior simulation.
|
| - **Success story:** A retailer can test churn, markdown and inventory policies against customer-behavior drift before margin loss reaches the P&L.
|
| - **Download:** `data/industry_vertical_suite/retail/semanta_retail_world_native_synthetic_100000x100.csv.gz`.
|
| - **Evidence:** `metrics/industry_vertical_suite/retail_quality_metrics.json` and `schemas/industry_vertical_suite/retail_schema.json`.
|
| - **Shape:** 100,000 rows x 100 columns; quality score 0.957.
|
|
|
| ## Example Success Stories
|
|
|
| - **Banks**: A bank can stress-test credit and fraud models against synthetic regime shifts before the same pattern appears in production.
|
| - **AI**: An AI team can generate controlled evaluation/training substrates for agents and models while keeping customer data out of external model providers.
|
| - **Hedge Funds**: A quant team can evaluate strategy fragility across liquidity shocks and factor crowding without waiting for the next real drawdown.
|
| - **Insurance**: An insurer can rehearse reserve and claim workflows against rare loss clusters before portfolio stress becomes visible in historical data.
|
| - **Pharma**: A pharma team can simulate trial-operation bottlenecks and adverse-event signal drift before allocating expensive clinical capacity.
|
| - **Medicine**: A hospital can evaluate triage and readmission models under capacity shock without exposing patient data to external systems.
|
| - **Manufacturing**: A plant can train predictive quality and downtime models on rare failure pathways without breaking real production equipment.
|
| - **Security**: A security team can harden detection pipelines against adversarial behavior and alert floods before analysts are overloaded in production.
|
| - **Science**: A lab can prioritize experiments by testing hypothesis robustness and replication risk before spending scarce lab time.
|
| - **FMCG**: An FMCG team can rehearse promotion and channel-mix decisions across demand shocks before committing trade-spend.
|
| - **Retail**: A retailer can test churn, markdown and inventory policies against customer-behavior drift before margin loss reaches the P&L.
|
|
|
| ## Files
|
|
|
| - `banks`: `data/industry_vertical_suite/banks/semanta_banks_world_native_synthetic_100000x100.csv.gz` | metrics `metrics/industry_vertical_suite/banks_quality_metrics.json` | schema `schemas/industry_vertical_suite/banks_schema.json`
|
| - `ai`: `data/industry_vertical_suite/ai/semanta_ai_world_native_synthetic_100000x100.csv.gz` | metrics `metrics/industry_vertical_suite/ai_quality_metrics.json` | schema `schemas/industry_vertical_suite/ai_schema.json`
|
| - `hedge_funds`: `data/industry_vertical_suite/hedge_funds/semanta_hedge_funds_world_native_synthetic_100000x100.csv.gz` | metrics `metrics/industry_vertical_suite/hedge_funds_quality_metrics.json` | schema `schemas/industry_vertical_suite/hedge_funds_schema.json`
|
| - `insurance`: `data/industry_vertical_suite/insurance/semanta_insurance_world_native_synthetic_100000x100.csv.gz` | metrics `metrics/industry_vertical_suite/insurance_quality_metrics.json` | schema `schemas/industry_vertical_suite/insurance_schema.json`
|
| - `pharma`: `data/industry_vertical_suite/pharma/semanta_pharma_world_native_synthetic_100000x100.csv.gz` | metrics `metrics/industry_vertical_suite/pharma_quality_metrics.json` | schema `schemas/industry_vertical_suite/pharma_schema.json`
|
| - `medicine`: `data/industry_vertical_suite/medicine/semanta_medicine_world_native_synthetic_100000x100.csv.gz` | metrics `metrics/industry_vertical_suite/medicine_quality_metrics.json` | schema `schemas/industry_vertical_suite/medicine_schema.json`
|
| - `manufacturing`: `data/industry_vertical_suite/manufacturing/semanta_manufacturing_world_native_synthetic_100000x100.csv.gz` | metrics `metrics/industry_vertical_suite/manufacturing_quality_metrics.json` | schema `schemas/industry_vertical_suite/manufacturing_schema.json`
|
| - `security`: `data/industry_vertical_suite/security/semanta_security_world_native_synthetic_100000x100.csv.gz` | metrics `metrics/industry_vertical_suite/security_quality_metrics.json` | schema `schemas/industry_vertical_suite/security_schema.json`
|
| - `science`: `data/industry_vertical_suite/science/semanta_science_world_native_synthetic_100000x100.csv.gz` | metrics `metrics/industry_vertical_suite/science_quality_metrics.json` | schema `schemas/industry_vertical_suite/science_schema.json`
|
| - `fmcg`: `data/industry_vertical_suite/fmcg/semanta_fmcg_world_native_synthetic_100000x100.csv.gz` | metrics `metrics/industry_vertical_suite/fmcg_quality_metrics.json` | schema `schemas/industry_vertical_suite/fmcg_schema.json`
|
| - `retail`: `data/industry_vertical_suite/retail/semanta_retail_world_native_synthetic_100000x100.csv.gz` | metrics `metrics/industry_vertical_suite/retail_quality_metrics.json` | schema `schemas/industry_vertical_suite/retail_schema.json`
|
|
|
| ## Quick Load Example
|
|
|
| ```python
|
| import pandas as pd
|
|
|
| url = "https://huggingface.co/datasets/SemantaAI/semanta-dataset-suite/resolve/semanta/data/industry_vertical_suite/banks/semanta_banks_world_native_synthetic_100000x100.csv.gz"
|
| df = pd.read_csv(url)
|
| print(df.shape)
|
| print(df.head())
|
| ```
|
|
|
| ## Load Any Vertical
|
|
|
| ```python
|
| import pandas as pd
|
|
|
| base = "https://huggingface.co/datasets/SemantaAI/semanta-dataset-suite/resolve/semanta"
|
| vertical = "ai" # banks, hedge_funds, insurance, pharma, medicine, manufacturing, security, science, fmcg, retail, ai
|
| url = f"{base}/data/industry_vertical_suite/{vertical}/semanta_{vertical}_world_native_synthetic_100000x100.csv.gz"
|
| df = pd.read_csv(url)
|
| print(vertical, df.shape)
|
| ```
|
|
|
| ## Quality Gates
|
|
|
| - `synthetic_only`: **pass** - No customer/private source data is used.
|
| - `per_industry_files`: **pass** - Each vertical has its own CSV, schema and metrics.
|
| - `100_column_width`: **pass** - Each vertical has 100 columns.
|
| - `scenario_coverage`: **pass** - All canonical scenarios are represented.
|
| - `hf_interop`: **pass** - CSV.GZ + README + schema JSON + metrics JSON.
|
| - `claim_boundary`: **pass** - No model-performance claims without benchmarks.
|
| - `production_smoke_gate`: **pass** - 13/13 public proof targets pass in the Semanta production smoke pipeline.
|
|
|
| ## Metrics Published
|
|
|
| | Metric family | Included evidence |
|
| |---|---|
|
| | Shape | rows, columns, cells and per-industry schema width |
|
| | Quality | missing rate, duplicate rate, target-label profile and overall quality score |
|
| | Scenario coverage | regimes, shocks, rare events, adversarial and recovery paths |
|
| | Reproducibility | seed, package manifest, lineage and deterministic generation contract |
|
| | Safety | synthetic-only policy, no customer/private source data and no external model provider use |
|
|
|
| ## How Semanta Generated This
|
|
|
| 1. Semanta selects an industry world specification: buyer, use case, domain signals and risk profile.
|
| 2. It samples canonical scenario regimes: baseline, growth, efficiency push, drift, stress, shock, rare event, adversarial and recovery.
|
| 3. It creates common world features shared across industries: macro, risk, latency, quality, drift, tail-risk and confidence signals.
|
| 4. It creates industry-native features for each vertical, such as credit risk for banks or trial signal for pharma.
|
| 5. It derives target labels and risk scores from scenario pressure, domain bias, anomaly density and drift.
|
| 6. It writes HF-compatible CSV.GZ files plus schemas, metrics, dataset index, lineage and claim boundaries.
|
|
|
| No customer data was used. No customer data was sent to DeepSeek, Gamma or any external model provider. This package is
|
| deterministic and reproducible from the recorded seed and Semanta generation contracts.
|
|
|
| ## Recommended Uses
|
|
|
| - Tabular ML classification and regression.
|
| - Scenario robustness testing.
|
| - Synthetic data quality experiments.
|
| - Drift and anomaly detection prototypes.
|
| - StarForge/Gamma training substrate preparation after downstream evaluation.
|
| - Investor/client demos showing Semanta's breadth across industries.
|
|
|
| ## Semanta Links
|
|
|
| - Website: https://semanta.xyz
|
| - Public API proof: https://api-staging.semanta.xyz/public/industry-dataset-suite
|
| - Public proof bundle: https://api-staging.semanta.xyz/public/proof-bundle?rows=64&seed=37
|
| - Studio Guided Demo: https://studio-staging.semanta.xyz
|
| - API docs: https://api-staging.semanta.xyz/docs
|
| - Operator contact: operator@semanta.xyz
|
|
|
| ## Good-Fit Use Cases
|
|
|
| | Buyer | Example use |
|
| |---|---|
|
| | AI teams | Stress-test agents and classifiers on synthetic task worlds before production traffic. |
|
| | Banks and fintechs | Explore credit, fraud, liquidity and policy-change scenarios without exposing customer data. |
|
| | Hedge funds and quants | Evaluate strategy fragility across synthetic market regimes and liquidity shocks. |
|
| | Healthcare and pharma teams | Prototype privacy-safe clinical, safety and trial-operation workflows. |
|
| | Manufacturing and security teams | Generate failure, sensor drift, attack and incident variants for robustness testing. |
|
|
|
| ## Claim Boundary
|
|
|
| This package proves Semanta's synthetic generation, packaging, metricing and HF delivery capability. It does **not**
|
| claim downstream model-performance superiority until StarForge/Gamma training and benchmark evidence is attached.
|
|
|
