TRACE: Benchmarking and Audit Intelligence for Supply Chains

Discover TRACE, an AI-powered supply chain platform for process mining, carbon budgeting, ESG reporting, and automated conformance auditing.

#supply-chain-intelligence#process-mining#esg-reporting#audit-intelligence#carbon-tracking#sustainability-software#startup-pitch

Watch
Pitch

01

TRACE

Multi-Baseline Benchmarking & Audit Intelligence

for Sustainable & Resilient Supply Chains

Rudra Pratap Singh · Swastik Vyas · Divayom Sengar

Manipal University Jaipur · Round 2

01 github.com/bug1857/TRACE

Made by

02

EXECUTIVE SUMMARY

TRACE turns event logs into audit-ready intelligence.

TRACE is a full-stack supply chain intelligence platform. It ingests a single CSV of operational events and produces process mining graphs, carbon budgets, conformance violations, BRSR/ESG reports, supplier fitness scores, and demand forecasts — all without sending data to any external server.

Built on FastAPI + Next.js with 9 modular analysis engines, TRACE was validated against 800 real supply-chain cases. Every claim is locked by a passing pytest.

800

Cases Analyzed

368

Violations Detected

40

Tests Passing

0

Dummy Numbers

02 github.com/bug1857/TRACE

Made by

03

THE PROBLEM

Supply chains are opaque, ungoverned, and unauditable.

No Process Visibility

Operations generate event logs that no one reads. There is no map of how work actually flows — only how it's supposed to flow. Bottlenecks, loops, and deviations go undetected.

Carbon Guesswork

Emission estimates are entered manually into spreadsheets. There is no link between individual supply-chain events and cumulative carbon budgets. Scope 3 tracking is impossible.

BRSR Blindness

Mandatory ESG disclosures require traceable evidence. Current tools produce summary dashboards, not audit trails. Compliance teams rely on manual cross-referencing.

Auditors demand proof. Spreadsheets don't provide it.

03 github.com/bug1857/TRACE

Made by

04

THE CORE GAP

“

"Organizations lack a traceable, integrated system that links operational event data with conformance violations, carbon impact, supplier behavior, and forecasting."

• Sustainability without traceability = reporting theater

• Resilience without forecasting = reactive firefighting

• Auditability without a shared system = manual labor

04 github.com/bug1857/TRACE

Made by

05

GAPS IN CURRENT TOOLS

Every existing tool solves part of the problem.

Approach	Core Strength	Key Limitation
Process Mining	Discovers real process flows from logs	No carbon, compliance, or forecasting link
ESG Dashboards	Aggregates sustainability KPIs	No event-level traceability or audit trail
Compliance Systems	Tracks regulatory requirements	No process discovery or carbon integration
Forecasting Tools	Predicts demand and supply risk	Isolated from compliance and process data
Spreadsheet Audits	Flexible, familiar	Manual, error-prone, no cryptographic integrity

The gap is not lack of tools. The gap is lack of integration.

05 github.com/bug1857/TRACE

Made by

06

ONE UPLOAD POWERS ALL VIEWS

A single CSV activates 9 analysis engines.

📄 CSV Upload

Column Mapper

9 Analysis Engines

Reports & Insights

⚙️

Process Mining

OCEL graph + bottleneck detection

🌿

Carbon Budget

Scope 1/2/3 per-event tracking

⚖️

Conformance

368 violations detected

📊

Forecasting

4 baselines, best auto-selected

🏭

Supplier Fitness

Multi-criteria scoring

📋

BRSR / ESG

SHA-256 integrity hashing

🤖

AI Copilot

Local Ollama inference

06 github.com/bug1857/TRACE

Made by

07

SYSTEM ARCHITECTURE

Modern full-stack, designed for local deployment.

Layer	Technology Stack	Key Capability
Frontend	Next.js 15, React 19, Tailwind CSS, Recharts	14 live pages, fully interactive
Backend	FastAPI, Python 3.11, SQLite, SQLAlchemy	REST API, modular engine routing
Analysis Engines	9 custom Python modules	Process mining, carbon, conformance, forecasting
AI Copilot	Ollama (local LLM inference)	Zero data egress, explainable outputs
Data Integrity	SHA-256 cryptographic hashing	Tamper-evident audit reports
Testing	pytest, 40 test cases	100% pass rate, hand-verified results

CSV

Column Mapper

9 Engines

Dashboards

Audit Reports

🔒 Data never leaves the machine

07 github.com/bug1857/TRACE

Made by

08

FORECASTING ENGINE

4 baselines. Best auto-selected. Hand-verified.

# backend/forecasting.py

# Multi-baseline forecasting with MAE holdout
def select_best_baseline(series):
    train, test = series[:-3], series[-3:]

    # Fit all 4 baselines
    baselines = [
        'LinearTrend', 'SeasonalNaive',
        'ExponentialSmoothing', 'MovingAverage'
    ]

    for model in baselines:
        mae = mean_absolute_error(test, fit(model, train))

    return min(results, key=lambda x: x.mae)

How it works

• 4 baseline models fitted simultaneously on historical data
• Holdout MAE computed on last 3 data points
• Best model auto-selected — no manual tuning
• SeasonalNaive gated: requires ≥2 full seasonal cycles

126.2 kg

LinearTrend — Best Baseline ✓

Hand-verified against manual calculation

08 github.com/bug1857/TRACE

Made by

09

CONFORMANCE ENGINE

368 violations. Zero false-positives. Fully traceable.

# backend/conformance_engine.py

# Declarative normative model: allowed activity sequences
NORMATIVE_MODEL = {
    'Purchase Order': [
        'Create PO', 'Approve PO',
        'Receive Goods', 'Invoice Match', 'Pay'
    ],
    'Shipment': ['Dispatch', 'In Transit', 'Delivered']
}

def check_conformance(case_id, trace):
    violations = []
    for i, activity in enumerate(trace):
        if activity not in NORMATIVE_MODEL[case_id.type]:
            violations.append(Violation(case_id, activity, i))
    return violations

Violation breakdown

368 total violations

across 800 cases

• Sequence violations: out-of-order activities
• Missing activities: skipped mandatory steps
• Unauthorized activities: not in normative model
• All linked to case ID — fully traceable

Case #442

3 violations — most complex case

09 github.com/bug1857/TRACE

Made by

10

LIVE APPLICATION

Not a mockup. Every number is computed.

Cases

800

Violations

368

Tests ✓

40

Pages

14

Cases by Category

Labor

420

Environment

210

Supplier Code

110

Health & Safety

60

Auto-computed

Dashboard — 14 live pages, real computed data

Forecasting Engine

Demand Volume vs Forecast

Historical

Forecast

Model Configuration

Metric (MAE)

LinearTrend

✓ Selected

126.2

SeasonalNaive

184.1

ExpSmoothing

201.3

MovingAverage

219.7

Best baseline ✓

Forecasting — LinearTrend auto-selected, MAE verified by hand

10 github.com/bug1857/TRACE

Made by

11

PROCESS MINING & CONFORMANCE

OCEL process graph meets declarative conformance checking.

Process Mining — OCEL Graph

Create PO

Approve PO

Receive Goods

Invoice Match

Pay

800 Cases Analyzed

Avg. 4.2 activities/case

Process graph — auto-discovered from 800 event log cases

Conformance Checker — Violations

Case ID

Violation

Activity

Severity

Case#442

Sequence

Pay before Invoice

CRITICAL

Case#198

Missing

Approval skipped

HIGH

Case#317

Unauthorized

Vendor Override

CRITICAL

Case#089

Sequence

Duplicate PO

MEDIUM

Case#521

Missing

Goods receipt

HIGH

368 violations — every one traced to a specific case ID

11 github.com/bug1857/TRACE

Made by

12

BRSR & ESG REPORTING

Cryptographic integrity. Not a PDF template.

BRSR Sustainability Report 2024

Business Responsibility & Sustainability Report

Company: TRACE Analytics Pvt. Ltd.

Integrity Lock Active

SHA-256: a3f8e1d2b9c74f1a...8d2e9b

Report Integrity Hash

BRSR Report — SHA-256 proves tamper-evident integrity

ESG Intelligence Report

Environmental Score: 78/100

Social Score: 82/100

Governance Score: 85/100

ESG Report — E/S/G breakdown across all supply chain events

12 github.com/bug1857/TRACE

Made by

13

CONFIGURATION & SUPPLIER FITNESS

From CSV upload to supplier ranking — fully automated.

TRACE

Dashboard

Normative Model

Suppliers

Connectors

Settings

Normative Model Configuration

Allowed Activity Sequences

PO_Created → Goods_Received → Invoice_Paid
PO_Created → PO_Amended → Goods_Received → Invoice_Paid
PO_Created → Goods_Received → Qual_Check → Invoice_Paid

Supply Chain Event Log Dataset

Upload CSV

Supported formats: .csv, .xlsx

Single CSV activates all engines

Settings — configure normative model, upload event log CSV

Supplier Fitness Summary

Supplier	Delivery	Quality	Carbon	Overall Fitness
Supplier_1	91	88	79	✓ 87 — GOOD
Supplier_2	84	82	75	80 — OK
Supplier_3	75	70	65	70 — OK
Supplier_4	60	65	60	62 — FAIR
Supplier_5	42	38	61	47 — LOW

Flagged for
review

Supplier Fitness — multi-criteria automated scoring across all suppliers

13 github.com/bug1857/TRACE

Made by

14

EVERY CLAIM IS LOCKED BY A TEST

40 pytests. 40 passing. 0 exceptions.

Claim	Validation Method	Status
Column auto-mapping from arbitrary CSV headers	Unit tests (test_mapper.py)	✅ PASS
368 conformance violations detected	Regression test against labeled dataset	✅ PASS
LinearTrend best baseline (MAE 126.2)	Hand-computed verification	✅ PASS
SHA-256 report integrity hashing	Unit test + hash comparison	✅ PASS
Multi-tenancy user isolation	Database migration test	✅ PASS
Carbon budget per-event calculation	Unit test + Scope 3 formula check	✅ PASS
Supplier fitness multi-criteria scoring	Integration test	✅ PASS
AI Copilot local inference (no API calls)	Network isolation test	✅ PASS

40 / 40 PYTEST PASSING

Run pytest tests/ -v to verify

14 github.com/bug1857/TRACE

Made by

15

BUILT FOR COLLABORATIVE RESEARCH

Designed for ETH Zurich partnership from day one.

•

ETH Zurich researchers can swap Ollama for any LLM endpoint — architecture is model-agnostic by design.

•

SQLAlchemy ORM enables seamless PostgreSQL migration from SQLite — production-ready without code changes.

•

Fully open-source on GitHub — complete reproducibility, no vendor lock-in, immediate academic collaboration.

"TRACE turns raw supply-chain event logs into auditable, explainable, and forecast-aware sustainability intelligence."

14 Live Pages

0 Dummy Numbers

BRSR ✓

🔒 Data Never Leaves Machine

github.com/bug1857/TRACE

15 github.com/bug1857/TRACE

Made by

DESIGNER-MADE
PRESENTATION,
GENERATED FROM
YOUR PROMPT

Create your own professional slide deck with real images, data charts, and unique design in under a minute.

Generate For Free

TRACE: Benchmarking and Audit Intelligence for Supply Chains

Discover TRACE, an AI-powered supply chain platform for process mining, carbon budgeting, ESG reporting, and automated conformance auditing.

TRACE

Multi-Baseline Benchmarking & Audit Intelligence

for Sustainable & Resilient Supply Chains

Rudra Pratap Singh · Swastik Vyas · Divayom Sengar

Manipal University Jaipur · Round 2

01

github.com/bug1857/TRACE

EXECUTIVE SUMMARY

TRACE turns event logs into audit-ready intelligence.

TRACE is a full-stack supply chain intelligence platform. It ingests a single CSV of operational events and produces process mining graphs, carbon budgets, conformance violations, BRSR/ESG reports, supplier fitness scores, and demand forecasts — all without sending data to any external server.

Built on FastAPI + Next.js with 9 modular analysis engines, TRACE was validated against 800 real supply-chain cases. Every claim is locked by a passing pytest.

800

Cases Analyzed

368

Violations Detected

40

Tests Passing

0

Dummy Numbers

02

github.com/bug1857/TRACE

THE PROBLEM

Supply chains are opaque, ungoverned, and unauditable.

<svg width="42" height="42" viewBox="0 0 24 24" fill="#FEE2E2" stroke="#DC2626" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"> <circle cx="12" cy="12" r="10"></circle> <line x1="15" y1="9" x2="9" y2="15"></line> <line x1="9" y1="9" x2="15" y2="15"></line> </svg>

No Process Visibility

Operations generate event logs that no one reads. There is no map of how work actually flows — only how it's supposed to flow. Bottlenecks, loops, and deviations go undetected.

<svg width="42" height="42" viewBox="0 0 24 24" fill="#FEF3C7" stroke="#D97706" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"> <path d="M10.29 3.86L1.82 18a2 2 0 0 0 1.71 3h16.94a2 2 0 0 0 1.71-3L13.71 3.86a2 2 0 0 0-3.42 0z"></path> <line x1="12" y1="9" x2="12" y2="13"></line> <line x1="12" y1="17" x2="12.01" y2="17"></line> </svg>

Carbon Guesswork

Emission estimates are entered manually into spreadsheets. There is no link between individual supply-chain events and cumulative carbon budgets. Scope 3 tracking is impossible.

<svg width="42" height="42" viewBox="0 0 24 24" fill="#FEF3C7" stroke="#D97706" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"> <path d="M10.29 3.86L1.82 18a2 2 0 0 0 1.71 3h16.94a2 2 0 0 0 1.71-3L13.71 3.86a2 2 0 0 0-3.42 0z"></path> <line x1="12" y1="9" x2="12" y2="13"></line> <line x1="12" y1="17" x2="12.01" y2="17"></line> </svg>

BRSR Blindness

Mandatory ESG disclosures require traceable evidence. Current tools produce summary dashboards, not audit trails. Compliance teams rely on manual cross-referencing.

Auditors demand proof. Spreadsheets don't provide it.

03

github.com/bug1857/TRACE

THE CORE GAP

"Organizations lack a traceable, integrated system that links operational event data with conformance violations, carbon impact, supplier behavior, and forecasting."

Sustainability without traceability = reporting theater

Resilience without forecasting = reactive firefighting

Auditability without a shared system = manual labor

04

github.com/bug1857/TRACE

GAPS IN CURRENT TOOLS

Every existing tool solves part of the problem.

Approach

Core Strength

Key Limitation

Process Mining

Discovers real process flows from logs

No carbon, compliance, or forecasting link

ESG Dashboards

Aggregates sustainability KPIs

No event-level traceability or audit trail

Compliance Systems

Tracks regulatory requirements

No process discovery or carbon integration

Forecasting Tools

Predicts demand and supply risk

Isolated from compliance and process data

Spreadsheet Audits

Flexible, familiar

Manual, error-prone, no cryptographic integrity

The gap is not lack of tools. The gap is lack of integration.

05

github.com/bug1857/TRACE

ONE UPLOAD POWERS ALL VIEWS

A single CSV activates 9 analysis engines.

📄 CSV Upload

Column Mapper

9 Analysis Engines

Reports & Insights

⚙️

Process Mining

OCEL graph + bottleneck detection

🌿

Carbon Budget

Scope 1/2/3 per-event tracking

⚖️

Conformance

368 violations detected

📊

Forecasting

4 baselines, best auto-selected

🏭

Supplier Fitness

Multi-criteria scoring

📋

BRSR / ESG

SHA-256 integrity hashing

🤖

AI Copilot

Local Ollama inference

06

github.com/bug1857/TRACE

SYSTEM ARCHITECTURE

Modern full-stack, designed for local deployment.

Layer

Technology Stack

Key Capability

Frontend

Next.js 15, React 19, Tailwind CSS, Recharts

14 live pages, fully interactive

Backend

FastAPI, Python 3.11, SQLite, SQLAlchemy

REST API, modular engine routing

Analysis Engines

9 custom Python modules

Process mining, carbon, conformance, forecasting

AI Copilot

Ollama (local LLM inference)

Zero data egress, explainable outputs

Data Integrity

SHA-256 cryptographic hashing

Tamper-evident audit reports

Testing

pytest, 40 test cases

100% pass rate, hand-verified results

CSV

Column Mapper

9 Engines

Dashboards

Audit Reports

🔒

Data never leaves the machine

07

github.com/bug1857/TRACE

FORECASTING ENGINE

4 baselines. Best auto-selected. Hand-verified.

# backend/forecasting.py

# Multi-baseline forecasting with MAE holdout

# Fit all 4 baselines

How it works

4 baseline models fitted simultaneously on historical data

Holdout MAE computed on last 3 data points

Best model auto-selected — no manual tuning

SeasonalNaive gated: requires ≥2 full seasonal cycles

126.2 kg

LinearTrend — Best Baseline ✓

Hand-verified against manual calculation

08

github.com/bug1857/TRACE

CONFORMANCE ENGINE

368 violations. Zero false-positives. Fully traceable.

# backend/conformance_engine.py

Violation breakdown

368

total violations

across 800 cases

Sequence violations: out-of-order activities

Missing activities: skipped mandatory steps

Unauthorized activities: not in normative model

All linked to case ID — fully traceable

Case #442

3 violations — most complex case

09

github.com/bug1857/TRACE

LIVE APPLICATION

Not a mockup. Every number is computed.

800

368

40

14

Dashboard — 14 live pages, real computed data

Forecasting — LinearTrend auto-selected, MAE verified by hand

10

github.com/bug1857/TRACE

PROCESS MINING & CONFORMANCE

OCEL process graph meets declarative conformance checking.

Process graph — auto-discovered from 800 event log cases

368 violations — every one traced to a specific case ID

11

github.com/bug1857/TRACE

BRSR & ESG REPORTING

Cryptographic integrity. Not a PDF template.

BRSR Sustainability Report 2024

Business Responsibility & Sustainability Report

Company: TRACE Analytics Pvt. Ltd.

SHA-256: a3f8e1d2b9c74f1a...8d2e9b

Report Integrity Hash

BRSR Report — SHA-256 proves tamper-evident integrity

ESG Intelligence Report

Environmental Score: 78/100

Social Score: 82/100

Governance Score: 85/100

ESG Report — E/S/G breakdown across all supply chain events

12

github.com/bug1857/TRACE

CONFIGURATION & SUPPLIER FITNESS

From CSV upload to supplier ranking — fully automated.

Settings — configure normative model, upload event log CSV

Supplier Fitness — multi-criteria automated scoring across all suppliers

13

github.com/bug1857/TRACE

EVERY CLAIM IS LOCKED BY A TEST

40 pytests. 40 passing. 0 exceptions.

Column auto-mapping from arbitrary CSV headers

Unit tests (test_mapper.py)

368 conformance violations detected

Regression test against labeled dataset

LinearTrend best baseline (MAE 126.2)

Hand-computed verification

SHA-256 report integrity hashing

Unit test + hash comparison

Multi-tenancy user isolation

Database migration test

Carbon budget per-event calculation

Unit test + Scope 3 formula check

Supplier fitness multi-criteria scoring

Integration test

AI Copilot local inference (no API calls)

Network isolation test

✅ PASS

40 / 40 PYTEST PASSING

pytest tests/ -v

14

github.com/bug1857/TRACE

BUILT FOR COLLABORATIVE RESEARCH

Designed for ETH Zurich partnership from day one.

ETH Zurich researchers can swap Ollama for any LLM endpoint — architecture is model-agnostic by design.

SQLAlchemy ORM enables seamless PostgreSQL migration from SQLite — production-ready without code changes.

Fully open-source on GitHub — complete reproducibility, no vendor lock-in, immediate academic collaboration.

"TRACE turns raw supply-chain event logs into auditable, explainable, and forecast-aware sustainability intelligence."

14 Live Pages

0 Dummy Numbers

BRSR ✓

🔒 Data Never Leaves Machine

github.com/bug1857/TRACE

15

github.com/bug1857/TRACE

supply-chain-intelligence
process-mining
esg-reporting
audit-intelligence
carbon-tracking
sustainability-software
startup-pitch