1. Executive Summary

Metalytics is a job-work and material accountability platform built for manufacturing MSMEs that send raw material to job workers for plating, polishing, finishing and similar processes – for example, the brass-parts industry of Jamnagar.

In these units, every movement of material is recorded on a handwritten slip called a chitthi. Metalytics keeps that habit but makes it reliable: the storekeeper writes the chitthi on a printed Metalytics pad, takes one photo, and on-device AI reads every box and creates a verified entry. From that point, every kilogram is tracked out and back – with loss (ghati) measured, the Section 143 one-year limit monitored, and Rule 55 challans, ITC-04 and audit evidence produced automatically.

Unlike general accounting software, which tracks money, Metalytics tracks the metal: what is lying with which job worker, for how long, and how much was lost. It runs entirely on the client’s own computer, with no AI subscription or API cost, and works in English, Gujarati and Hindi.

Metalytics on desktop and phone – live 3D map of the works and job workers

2. Business Problem

Manufacturing MSMEs routinely send material out for job work, but the record of that material lives on paper. Typical workflow:

1. Material weighed and sent out on a handwritten chitthi
2. Returns (aavak) noted on another slip, often without reference to the original chitthi
3. Process loss (ghati) estimated, not measured
4. Stock with job workers “confirmed” from memory at year-end
5. ITC-04 prepared at the last minute from a pile of slips

Challenges

1. No visibility: the owner cannot say how much material is lying with each job worker, or since when
2. Hidden losses: normal process loss and genuinely missing material look the same on paper
3. GST exposure: inputs not returned within one year are deemed supplied under Section 143 of the CGST Act, with tax and interest – and nobody is counting the days
4. Compliance effort: Rule 55 delivery challans, e-way bills for inter-State job work and ITC-04 are prepared manually
5. Audit difficulty: CARO 2020 clause 3(ii)(a) requires reporting on physical verification of inventory, including stock with third parties – which cannot be verified from paper slips
6. No controls: shortages can be written off without a second person’s review, and there is no trail of changes
7. Language barrier: storekeepers often work in Gujarati or Hindi; most ERPs are English-only and costly

3. Solution Overview

Metalytics digitises the job-work cycle without changing how the factory works. Its core idea is to keep the chitthi and make it smart.

1. Chitthi pad + on-device AI reading: a printed form with corner markers, digit boxes and tick boxes, read box by box from a phone photo
2. Material ledger by kilogram: every chitthi, return, ghati and shortage tracked at the job-worker level
3. Compliance engine: Section 143 ageing, Rule 55 challans, e-way bill checks, ITC-04 tables and review
4. Controls: role-based rights for five roles, maker–checker approvals and a tamper-evident change log
5. Insight: a live 3D map, analytical reports and natural-language questions in three languages

Reading a chitthi pad photo – each figure shown beside the part of the photo it came from

4. End-to-End Workflow

Step 1 – Order entry

The order line is recorded – party, size, finish, quantity and weight per dozen. The expected weight is calculated automatically.

Step 2 – Print the chitthi pad

A Metalytics pad is printed with the current job workers, waiting order lines and peti (container) types, and a machine-readable pad number.

Step 3 – Write and photograph

The storekeeper writes the date, peti count and gross weight in digit boxes and ticks the job worker, order line and peti type, then takes one photo.

Step 4 – AI reading on the device

Metalytics locates the four corner markers, straightens the page, reads ticks by measuring ink and reads digits with an on-device neural network. Each value is shown next to its crop of the photo.

Step 5 – Validation and confirmation

Weight is compared with the order, dates are checked for sense, and uncertain figures are highlighted. Nothing is saved until a person confirms.

Step 6 – Delivery challan

A Rule 55 delivery challan is generated in English, Gujarati or Hindi, with amount in words, declaration and signatures. E-way bill is enforced for inter-State job work.

Step 7 – Return (aavak) and ghati

Returns are weighed in against the job number. Ghati is measured against the agreed tolerance, and any unexplained difference is routed to a second person for approval.

Step 8 – Monitoring and alerts

A live 3D map and the “At finishing” screen show material outside, its age and value. Warnings begin at 270 days, before the one-year limit.

Step 9 – Compliance and audit outputs

ITC-04 Table 4 and 5A are prepared and checked, an audit memo lists exceptions, and job-worker statements support external confirmation.

Step 10 – Reports and questions

Analytical reports (scorecards, ageing, trends, loss) are exported as designed PDFs or Excel, and users can ask questions in English, Gujarati or Hindi.

5. Major Features

6. How AI Is Used

1. Computer vision: corner detection and perspective correction straighten any photo of the pad, even tilted or upside down
2. Handwriting recognition: a convolutional neural network, trained on about 60,000 handwritten digits and adapted to Indian writing styles, reads each digit on the device with a confidence score
3. OCR: Tesseract reads free-hand chitthis in English, Gujarati and Hindi
4. Natural-language questions: an intent engine understands questions in three languages, including Gujarati word endings, and answers only from the client’s data
5. AI-assisted development: the application was designed by a Chartered Accountant and built using Claude Code, an AI coding assistant

Deterministic rules – not AI – are used for every tax, stock and money calculation, so the same data always gives the same result.

7. Compliance Coverage

8. System Architecture

1. Runs entirely in the browser on the client’s computer; no server holds client data
2. Data is saved in the browser, with an optional automatic copy to a folder and a one-click backup file
3. Works offline after the first load, including photo reading

9. Technology Stack

10. Data Privacy & Security

1. No client data is sent to any AI company or external server; even the photo is read on the device
2. No API key, subscription or per-use cost
3. Five roles (Owner, Accountant, Supervisor, Storekeeper, CA) with specific rights; the CA login is read-only
4. Every change is recorded with who, when and what it was before; the log cannot be switched off or edited

11. Business Impact

1. Real-time visibility of material lying with every job worker
2. Loss identified lot by lot, separating normal ghati from shortages
3. Reduced GST exposure from goods crossing the one-year limit
4. ITC-04 and challans prepared from source records, not reconstructed
5. Audit-ready evidence for inventory with third parties
6. Minimal change for floor staff – the chitthi stays, in their own language
7. Zero running cost and full data privacy

Section 143 monitoring – material with job workers by age, value and status

12. Future Roadmap

1. Shared multi-user server deployment with secure login
2. Tally integration for vouchers and stock
3. Capital goods three-year limit and ITC-04 Tables 5B and 5C
4. WhatsApp alerts for overdue material and approvals
5. Multi-location and multi-company support
6. Optional cloud AI reading for free-hand slips, with client consent

13. Conclusion

Metalytics turns the humble chitthi into a verified, auditable record. By combining on-device AI with the domain knowledge of a Chartered Accountant, it gives manufacturing MSMEs control over the material they send for job work, helps them stay compliant with GST requirements, and gives auditors reliable evidence – without changing how the factory works and without any data leaving the business.