A living proposal — not a PDF

From tide to table, every gram accounted.

Wisewit brings AI-powered weighing intelligence to Nekkanti Seafoods — automated capture from the unloading bay to the beheading hall, with OCR, computer vision, and reconciliation math we are prepared to stand behind.

Phase 1Automated Weighing

StackOCR · Computer Vision · Edge AI

PreparedJuly 2026 · Confidential

The water remembers everything.WW×NSF/26

Why this proposal is different

Three commitments, numbered like a manifesto.

CHAPTER 01 — UNIQUENESS

A proposal that behaves like the product.

This page is live. The questions have answer boxes. The commercials recalculate as you read. The plant layout is a working schematic. We didn't attach a document — we shipped a small piece of software, because that is what we do.

CHAPTER 02 — STRENGTH

OCR + Computer Vision, edge-first.

Camera-read weighments, crate counting, material-count estimation, ANPR at the gate, and an edge architecture designed for near-100% continuity of capture — even when the network is not.

CHAPTER 03 — COMMITMENT

Math we are willing to sign.

Raw material in ≈ beheaded yield + wastage out. Every batch, every supplier, every shift — reconciled automatically, with variance flagged before it becomes a dispute. That is our promise on deliverables.

35+

Years — Nekkanti in business

$300MM

Annual revenue

100MM lbs

Sold annually

100+

Customers worldwide

The scale we are building for — India's largest fully integrated shrimp player.

Deliverables timeline

Kickoff to go-live in 4–5 weeks.

Stage 1

GPU & Edge Hardware Procurement, Setup & Commissioning

2 – 3 Weeks

Stage 2

AI Model Integration & On-site Pipeline Deployment

2 Weeks

Stage 3

On-site Fine-Tuning, Calibration & User Acceptance Testing (UAT)

1 Week

Total

Project Kickoff to Full Go-Live

4 – 5 Weeks

KICKOFF · August 20thSTAGE 1 · HARDWARESTAGE 2 · AI PIPELINESTAGE 3 · UATGO-LIVE · September 24th

Timeline example — if we initiate the project on August 20th, the target go-live date will be September 24th.

02 — First-Step Solution

Automated weighing, from the truck to the yield.

Phase 1 records every weighment automatically — raw material at unloading, beheaded shrimp and wastage after beheading — and reconciles them per batch, per supplier, in real time.

Truck arrival & gross weighment

Supplier truck reaches the unloading area. The scale's LED display is read by CCTV via OCR — no manual entry. ANPR links the vehicle to the supplier.

Net raw material, per batch & supplier

Tare is deducted automatically. The net shrimp weight is logged against a Batch ID and Supplier ID, with timestamp and video evidence.

Crate counting

Computer vision counts crates unloaded per batch — declared vs. actual, cross-checked instantly.

Material count, at the table

Shrimp is spread in a crate on the count table. An overhead CCTV estimates the count per kilo — 30-count means 30 pieces per kg — replacing manual sampling.

Beheading

Heads and legs are removed in the beheading hall. The batch identity travels with the material.

Yield & wastage weighing

Beheaded shrimp is weighed; heads and legs (wastage) are weighed separately — both captured automatically against the same Batch ID.

The reconciliation math

What goes in must come out.

R ≈ B + W // raw material ≈ beheaded yield + wastage (heads & legs)
Δ = R − (B + W) // variance, flagged beyond ±0.5% tolerance
Yield% = B ÷ R × 100 // tracked per supplier, per variety, per shift

Raw material R (kg)

Beheaded B (kg)

Wastage W (kg)

Variance Δ / Yield

—

Interactive — type any figures. A live demo of what the floor dashboard shows, every batch.

Sample batch register

Every weighment, tied to a Batch ID.

Batch ID	Supplier	Vehicle	Raw (kg)	Beheaded (kg)	Wastage (kg)
NSF-2607-014	Surya Aqua Farms	AP09-TR-4412	1,240	798	431
NSF-2607-015	Godavari Marine	AP10-UT-9921	980	624	349
NSF-2607-016	Krishna Delta Seafoods	AP16-KP-1187	1,520	955	540
NSF-2607-017	Surya Aqua Farms	AP09-TR-4488	1,105	712	382

Δ and Yield compute live. Status: OK within ±0.5% · CHECK beyond tolerance. Admin can edit any cell.

03 — Questions from our end

Help us calibrate the first kilometre.

Answer right here in the boxes — everything you type auto-saves in this browser. Or reply over email; either way, these answers lock the Phase 1 bill of materials.

Answers auto-save locally & to the cloud · Nothing is shared until you choose to

0 / 14 ANSWERED0%

A — Weighing at unloading, from supplier trucks

GATE → UNLOADING BAY → SCALE

A1How many trucks and distinct suppliers arrive per day on average, and in what time windows?

A2Which platform scales are used at the unloading area — how many, and what make, model and capacity? Do they have RS-232 / Ethernet output, or only an LED display?

A3How is a "batch" defined today — per truck, per supplier, per lot? Can one truck carry multiple batches?

A4How are supplier names and vehicle numbers recorded today — register, Excel, or ERP? Who records them?

A5How is tare handled — per-crate tare at the platform scale, or gross and net in two passes? Are crates returned and reused across suppliers?

A6What variance (%) between supplier-declared weight and plant-received weight is acceptable today, and what happens when it is exceeded?

B — Weighing after the beheading area

BEHEADED SHRIMP + WASTAGE (HEADS & LEGS)

B1Where is the beheaded shrimp weighed — the same scale or a separate scale in the beheading hall? How many scales exist there?

B2Are heads and legs (wastage) collected and weighed per batch, or only aggregated per shift?

B3Is wastage sold, discarded, or processed further? Does its weight carry any commercial value that should be tracked?

B4Is there drip / water loss between unloading and beheading that should be captured as its own weight category?

B5Are partial batches re-weighed after grading or soaking, or weighed only once post-beheading?

C — The reconciliation math

RAW ≈ BEHEADED + WASTAGE

C1Is the check "raw ≈ beheaded + wastage" done manually today? At what frequency, and by whom?

C2What tolerance band (e.g., ±0.5%) should the system treat as acceptable before flagging a batch for investigation?

C3When variance exceeds tolerance, who should receive the alert — floor supervisor, QC head, or management? By what channel (dashboard, WhatsApp, email)?

04 — Our proposals

OCR & Computer Vision, mapped to your floor.

Every solution below is deployable on the same edge-first architecture. Admin can edit any cell or add rows.

#	Solution	Type	What it does	Value to Nekkanti
S1	Weigh-Scale OCR	OCR	CCTV reads the LED display of every scale; each weighment is auto-recorded with timestamp, Batch ID and supplier.	Zero manual entry. No transcription errors. Every kilo has video evidence.
S2	ANPR Gate Logging	OCR	Automatic number-plate recognition at the main gate links vehicles to suppliers and batches.	Truck-to-supplier-to-batch traceability without registers.
S3	Document OCR	OCR	Supplier challans, invoices and weighment slips are digitised into ERP-ready structured records.	Purchase data flows straight into accounts; no re-keying.
S4	Crate Counting	CV	Computer vision counts crates unloaded per batch, automatically.	Declared vs. actual crates cross-checked in real time.
S5	Material Count (count/kg)	CV	An overhead camera over the count table estimates pieces-per-kg from spread shrimp — e.g., 30-count = 30 pieces per kilo.	Automated count grading replaces slow, subjective manual sampling.
S6	Yield & Wastage Reconciliation	CV + Math	Per-batch reconciliation of raw vs. beheaded + wastage with tolerance alerts.	Leakage and disputes surface the same shift, not the same quarter.
S7	Size-Grading Vision	CV	Camera-based size distribution analysis on the grading line.	Consistent grading for export specifications.
S8	PPE & Hygiene Compliance	CV	Detects hairnets, gloves and aprons across processing halls.	Audit-ready compliance evidence for HACCP / BRC / USFDA.
S9	Quality Anomaly Flags	CV	Flags discoloration, melanosis signs and foreign bodies on inspection tables.	Early quality intervention before packing.
S10	Cold-Chain Telemetry	Telemetry	Line and cold-storage telemetry with alerting on the same platform.	One dashboard for yield, quality and temperature.

05 — Tech stack & configuration

Edge-first. Capture never stops.

Sized for 16–24 IP cameras plus scale and line telemetry. Final bill of materials locks once Section 03 answers arrive.

Layer	Component	Specification	Redundancy / Note
Capture	IP cameras	16–24 cameras, H.265, IP66 + vandal-rated for wet / outdoor zones	Positioned at gate, scales, count tables, halls
Capture	Weighing telemetry	Platform scale feed (RS-232 / Ethernet) + LED display OCR as an independent second channel	Dual-path capture: direct feed + camera proof
Edge compute	Edge server	Rack server, 8–16 cores, 32–64 GB RAM, on-site	Runs capture, local analytics, store-and-forward buffer
Edge compute	AI / GPU node	NVIDIA Blackwell 6000 Ada GPU node for on-premise inference	Events & clips sent to cloud, not raw 24×7 streams
Storage	Local storage	RAID 6 / RAID 10, enterprise + surveillance-grade drives	Hot spare; 30–90 days on-site retention
LAN	Switching	Managed L3 PoE+ core, L2 PoE+ access, VLAN segmentation, QoS, LACP / RSTP, NTP, redundant DNS	Redundant core; capture traffic prioritised
WAN	Connectivity	Primary Leased Line + 5G/4G failover + broadband secondary; SD-WAN / dual-WAN firewall; IPsec + TLS	Edge keeps capturing and buffering even if both links fail
Power	UPS + DG	Online double-conversion UPS; diesel generator with auto-start + ATS; dual PSU	No capture gap through power events
Cloud	Central platform	Geo-redundant region; ingestion, durable versioned storage, analytics, dashboards, reporting	Automatic backfill when links restore
Software	Platform services	Capture & ingestion · store-and-forward agent · buffer & retention manager · health agent · cloud validation · analytics / alerting · central config & OTA updates	Checksum + monotonic sequence on every record
Operations	Support model	24×7 monitoring, alerting, spares, preventive maintenance, SLA-governed response	Single accountable throat to choke: Wisewit

06 — Graphical plant layout

One batch's journey, under seven eyes.

Schematic of the Phase 1 capture points. Trucks enter the main gate and reach the unloading area directly — every weighing is read from the scale's LED display and recorded per Batch ID and per supplier. No weighbridge inside the plant.

Material flow (batch travels with its ID) CCTV capture point (C1–C7) Processing zone

Point	Location	Reads / Detects	Feeds
C1	Main gate	Number plate (ANPR), arrival time	Supplier & vehicle linking
C2	Unloading scale LED	Every weighment via OCR	Raw material weight per batch & supplier
C3	Unloading bay	Crate count	Declared vs. actual crates
C4	Overhead at count table	Material count per kg (e.g., 30-count)	Count/size grading record
C5	Beheading hall	Process coverage, batch continuity	Traceability evidence
C6	Yield scale A	Beheaded shrimp weight via OCR	Yield per batch
C7	Wastage scale B	Heads & legs weight via OCR	Wastage per batch · reconciliation

Camera schedule expands to 16–24 cameras as later phases add grading lines, PPE compliance and cold storage.

07 — Commercials

Three ways to engage. One recommendation.

All figures in ₹ (INR). Admin can edit every field, adjust the revenue-share percentage live, and add rows to the per-feature model.

OPTION 1

Revenue Share

Wisewit takes a share of the savings our system generates for Nekkanti — leakage prevented, disputes eliminated, yield recovered. If we don't save, we don't earn.

15%

Estimated monthly savings (₹)

Wisewit share / month

₹3,00,000

Wisewit share / year

₹36,00,000

Justification

+ One-time GPU: ₹25,00,000 (fixed)

Caution: Wisewit will source and commission the GPU, but Nekkanti pays at actuals as per the vendor agreement.

OPTION 2

Per Solution / Feature

Pay per solution after successful deployment and acceptance. Each feature is scoped, priced and justified individually in the table below.

Scoped per feature

Development + deployment cost per feature, plus monthly maintenance. Rows added by Admin as scope firms up.

Justification

+ One-time GPU: ₹25,00,000 (fixed)

Caution: Wisewit will source and commission the GPU, but Nekkanti pays at actuals as per the vendor agreement.

Wisewit Recommendation

OPTION 3

Fixed Monthly Partnership

A fixed monthly engagement covering Phase 1 and any software development requirements from Nekkanti's end.

₹3.5 Lakhs / month, all-inclusive

Covers AI features, integrations and enhancements — one predictable line item.

Why we recommend this

+ One-time GPU: ₹25,00,000 (fixed)

Caution: Wisewit will source and commission the GPU, but Nekkanti pays at actuals as per the vendor agreement.

Option 2 — detail

Per-feature pricing, row by row.

Project	Feature	Dev + Deployment	Maintenance / mo	Justification
Yield Intelligence — Phase 1	Weigh-scale OCR + automatic batch logging (unloading)	₹4,50,000	₹25,000	Removes manual entry; every weighment tied to Batch ID and supplier with video proof.
Yield Intelligence — Phase 1	Beheading yield + wastage reconciliation engine	₹3,50,000	₹20,000	Live raw vs. (beheaded + wastage) variance with tolerance alerts to supervisors.
Vision Suite	Crate counting (computer vision)	₹2,75,000	₹15,000	Declared vs. actual crate counts per batch; stops quiet leakage at the bay.
Vision Suite	Material count CV (count-per-kg at the table)	₹3,25,000	₹18,000	Automates count grading (e.g., 30-count); consistent, auditable, export-ready.

Hardware (cameras, edge server, networking, UPS) is billed at actuals under any model. Taxes extra as applicable.

08 — For the Sr. VP

44 questions, eight sections.

Ready to email. Copy them in one click, or share this page. Answers to these lock the final bill of materials, the SLA tier, and the Phase 1 go-live plan.

Answer right here — every answer auto-saves locally & to the cloud, and Wisewit can see the progress

0 / 44 ANSWERED0%

1 · Operations & Process Flow

6 QUESTIONS

1.1Walk us through a truck's journey today — gate → unloading → beheading → dispatch. Where are the bottlenecks?

1.2How many shifts run per day, and how many trucks are processed per shift?

1.3How many unloading bays exist, and can multiple trucks unload simultaneously?

1.4What is the average time from truck arrival to beheading completion for one batch?

1.5Are batches ever mixed — two suppliers in one lot — before beheading?

1.6When a discrepancy is found today, what happens — and who has authority to stop the line?

2 · Weighing, Yield & Wastage

7 QUESTIONS

2.1What scales exist at unloading and beheading today — make, model, capacity, and connectivity (RS-232 / Ethernet / Bluetooth)?

2.2Are the scale LED displays large and bright enough to be read by a camera from 3–5 metres?

2.3How often are scales calibrated, and by whom?

2.4What is the typical yield % from raw to beheaded for your main varieties?

2.5Is ice or water weighed together with the shrimp at any stage? How should the system treat it?

2.6Do you weigh wastage (heads and legs) per batch or per shift today?

2.7What variance tolerance (±%) is acceptable between raw material and beheaded + wastage?

3 · Suppliers & Procurement

5 QUESTIONS

3.1How many active suppliers deliver to the plant, and how are they coded or identified today?

3.2Do suppliers declare weights and counts on challans? How often are plant figures disputed?

3.3Is supplier payment linked to plant-received weight or supplier-declared weight?

3.4Do trucks ever arrive without documentation? How are they handled?

3.5Should suppliers get visibility of their own batch data — a receipt or a portal?

4 · Crates & Material Count

5 QUESTIONS

4.1What crate types and sizes are used, and are they barcoded or labelled?

4.2How is the material count (e.g., 30-count) determined today — manual sampling? How many pieces per sample?

4.3How many count tables exist, and is there space and lighting for an overhead camera above each?

4.4How often is the count re-verified after grading or soaking?

4.5Who records counts today, and where — register, Excel, or ERP?

5 · Infrastructure & Site Readiness

6 QUESTIONS

5.1Is there rack space, earthing, cooling and cable pathways for an edge server near the production floor?

5.2Is there a diesel generator with auto-start, and UPS coverage for network equipment?

5.3What is the current internet setup — leased line, broadband, and 4G/5G signal quality at the site?

5.4Are existing CCTV cameras IP-based, and can their feeds be shared with an analytics server?

5.5Are there restrictions on mounting cameras over count tables or scales — hygiene zones, washdown areas?

5.6Who manages on-site IT today, and can they support basic network tasks during installation?

6 · IT, ERP & Integration

6 QUESTIONS

6.1Which ERP / accounting system is in use — name and version? On-premise or cloud?

6.2Does the ERP expose APIs, database access, or file-based export for integration?

6.3Which modules should weighing data flow into — purchase, inventory, production, QC?

6.4Are there legacy or custom software systems that would need modification or re-writing to link with an AI platform?

6.5Who owns the ERP vendor relationship, and can the vendor be engaged for integration work?

6.6Are there data-residency or data-privacy requirements we should design for?

7 · Compliance, QC & Traceability

5 QUESTIONS

7.1Which certifications require traceability records — HACCP, BRC, USFDA, others — and for what retention period?

7.2What QC checks happen between unloading and beheading today?

7.3Are batch records audited by customers or regulators? In what format are they produced?

7.4Do export customers require count / size certificates per shipment?

7.5How are rejections and downgrades documented and traced back to supplier and batch?

8 · Commercials, Timeline & Governance

4 QUESTIONS

8.1What is the target go-live window for Phase 1, and are there production seasons we should avoid for installation?

8.2Who will be the single point of contact for approvals, and what does the sign-off process look like?

8.3Of the three commercial models — revenue share, per-feature, or fixed monthly — which structure does your finance team prefer?

8.4For the revenue-share model, what baseline savings figure should we use, and how is "savings" defined on your side?

09 — Solution architecture

Decoupled capture. Nothing lost, ever.

From the NSF_JTP Solution Architecture: capture is decoupled from delivery, so data capture continues near-100% even through network outages — edge first, cloud second, buffer always.

~100%

Continuity of data capture

0 gaps

For outages inside the buffer window

99.9%

SLA-backed cloud & dashboard availability

24×7

Monitoring & SLA-governed response

Layer 1 — Capture

Cameras & telemetry

IP cameras, scale feed, line and cold-storage telemetry at every capture point.

Layer 2 — Edge

Compute & storage

On-site edge server with GPU running capture, local analytics and the store-and-forward buffer on local RAID.

Layer 3 — LAN

Managed & segmented

PoE switching, redundant core, QoS that prioritises capture and telemetry traffic.

Layer 4 — WAN

Dual-path connectivity

Leased line primary, wireless failover, encrypted tunnels to cloud via SD-WAN / dual-WAN firewall.

Layer 5 — Cloud

Central platform

Geo-redundant ingestion, durable storage, analytics, dashboards and reporting.

Layer 6 — Ops

Support & SLA

24×7 monitoring, alerting, spares, preventive maintenance, SLA-governed response.

Availability reference

What the SLA numbers mean.

Availability	Downtime / year	Downtime / month	Position
99.0%	3.65 days	7.31 hours	Basic broadband-class service
99.9%	8.77 hours	43.8 minutes	Committed cloud & dashboard tier
99.95%	4.38 hours	21.9 minutes	Available upgrade tier
99.99%	52.6 minutes	4.38 minutes	Premium tier, on request

Domain	Redundancy measure
Internet path	Leased line + 5G/4G failover + broadband secondary; automatic failover at the WAN edge
LAN core	Redundant managed L3 core, LACP / RSTP link resilience
Edge compute	Spares on-site; health agent with remote monitoring
Storage	RAID 6 / RAID 10 + hot spare; geo-redundant versioned cloud copy
Power	Online double-conversion UPS + DG with ATS + dual PSU
Data integrity	Checksum + monotonic sequence ID on every record; reconciliation for any gap

Readiness: Section-0 inputs, retention & compliance requirements, cloud region and the final SLA tier lock once the Sr. VP questionnaire (Tab 08) is answered. Go / No-Go checklist follows.

10 — Phase 1 go-live timeline

From site survey to acceptance.

Indicative sequence for Phase 1. Dates are set by Wisewit admin — unlock Admin mode and click any date chip to enter it.

M1Site survey & inputs confirmation

Walk the gate, unloading bay, count tables and beheading hall; confirm scales, lighting, mounting points and Section-0 inputs.

TBD

M2BoM lock & commercial sign-off

Final bill of materials, SLA tier and commercial model agreed and signed by both parties.

TBD

M3GPU & hardware procurement

Wisewit sources and commissions the GPU node, cameras, edge server and networking — billed to Nekkanti at actuals per the vendor agreement.

TBD

M4Network, power & rack readiness

Leased line, VLANs, PoE switching, UPS and rack space provisioned at the plant.

TBD

M5Camera & scale installation

C1–C7 cameras mounted and aligned to scale LED displays; scale connectivity (RS-232 / Ethernet) tapped where available.

TBD

M6Edge node commissioning

Edge server live with capture, store-and-forward buffer and cloud sync; failover and buffer tests passed.

TBD

M7Model calibration

Scale OCR, crate counting and material-count models tuned on Nekkanti's real floor conditions, lighting and shrimp varieties.

TBD

M8Trial batches & UAT

Live batches run through the system; reconciliation accuracy validated against manual records with Nekkanti's team.

TBD

M9Go-live — Phase 1

Automated weighing and reconciliation live for all unloading and beheading weighments; dashboards handed over.

TBD

M10Acceptance & SLA start

Formal acceptance sign-off; 24×7 monitoring and SLA-governed support begin.

TBD

From tide to table, every gram accounted.

Three commitments, numbered like a manifesto.

A proposal that behaves like the product.

OCR + Computer Vision, edge-first.

Math we are willing to sign.

Kickoff to go-live in 4–5 weeks.

GPU & Edge Hardware Procurement, Setup & Commissioning

AI Model Integration & On-site Pipeline Deployment

On-site Fine-Tuning, Calibration & User Acceptance Testing (UAT)

Project Kickoff to Full Go-Live

Automated weighing, from the truck to the yield.

Truck arrival & gross weighment

Net raw material, per batch & supplier

Crate counting

Material count, at the table

Beheading

Yield & wastage weighing

What goes in must come out.

Every weighment, tied to a Batch ID.

Help us calibrate the first kilometre.

A — Weighing at unloading, from supplier trucks

B — Weighing after the beheading area

C — The reconciliation math

OCR & Computer Vision, mapped to your floor.

Edge-first. Capture never stops.

One batch's journey, under seven eyes.

Three ways to engage. One recommendation.

Revenue Share

Per Solution / Feature

Fixed Monthly Partnership

Per-feature pricing, row by row.

44 questions, eight sections.

1 · Operations & Process Flow

2 · Weighing, Yield & Wastage

3 · Suppliers & Procurement

4 · Crates & Material Count

5 · Infrastructure & Site Readiness

6 · IT, ERP & Integration

7 · Compliance, QC & Traceability

8 · Commercials, Timeline & Governance

Decoupled capture. Nothing lost, ever.

Cameras & telemetry

Compute & storage

Managed & segmented

Dual-path connectivity

Central platform

Support & SLA

What the SLA numbers mean.

From site survey to acceptance.

M1Site survey & inputs confirmation

M2BoM lock & commercial sign-off

M3GPU & hardware procurement

M4Network, power & rack readiness

M5Camera & scale installation

M6Edge node commissioning

M7Model calibration

M8Trial batches & UAT

M9Go-live — Phase 1

M10Acceptance & SLA start

Admin access