Skip to content
Backendz

Running systems, not slide decks

Dispatch boards, reconciliation agents, factory-floor apps and the ERP underneath. Every screen below is a build we shipped and still support.

ShiftTracker
ShiftTrackerShift management for security and FM teams.Backendz built ShiftTracker as one control room for every shift, shared by the three parties involved: the organisation that runs the rota, the agencies that supply officers and the clients who pay for the site. Each sees only what it needs. Compliance sits inside the workflow rather than in a report. Licence, double-booking and Working Time checks run at the moment a shift is assigned, when a breach can still be prevented. Time is tracked as scheduled, actual, expected and arbitrated, so disputes are settled once and payroll and client invoices use the same agreed hours. Lone workers are covered by scheduled check calls with PIN verification, geofencing and automatic escalation. Officers use the iOS and Android apps for rotas, check calls and 38 ready-made security and facilities forms, which work offline and produce PDFs on the client's letterhead.Web platformiOS appAndroid appLive in productionWays to fill an open shift4View case study
Digital Security Planner
Digital Security PlannerA site survey, design and estimating tool for security integrators.We built one tool for the whole job. Users upload a floor plan, then drag devices onto it from a library grouped by system: video surveillance, access control, intrusion detection, intercom, IT and infrastructure. Cameras show their field of view on the plan, so coverage gaps are visible before anything is ordered. Each device carries its own record: general details, equipment such as camera, mount, licence and accessories, installation specifics, notes, and site photos of the existing condition. From the placed devices the planner builds the bill of materials and labour estimate, then the proposal. Drawings are listed by customer, site, proposal number, design due date, priority, sales rep and sales engineer, and can be shared with installers, subcontractors and customers.React NativeNode.jsAWS Amplify~ Device categories in the library6View case study
Construction Site PPE & Safety Monitoring
Construction Site PPE & Safety MonitoringHard-hat and hi-vis checks, pedestrian safety-zone alerts and incident review from existing site cameras.We built a computer vision system that detects people, hard hats, hi-vis clothing and forklifts in the site camera feeds. PPE requirements are configured by area, so checks reflect the rules for the part of the site in view. Forklift detection adds a configurable buffer zone around the vehicle. When a person enters that zone, the system raises an alert for review. Each event includes a snapshot, camera, timestamp and duration, giving the supervisor the context needed to assess it. Events can be resolved or marked as false alarms rather than left as an unmanaged notification list. A start-of-shift view supports checks on the crew’s equipment, while weekly reporting groups compliance events by day and camera. This gives site managers a way to identify repeated issues and direct follow-up work. The system supports the site’s existing supervision and inspection routines; alerts remain subject to human review.AIComputer visionPython~ New cameras needed0View case study
Crowd Density Monitoring
Crowd Density MonitoringLive crowd counts, zone density and congestion alerts from existing CCTV, giving operations teams a clear view of where people are gathering.We built a computer vision pipeline that detects people in each camera view and assigns them to operator-defined zones using their position on the floor. Each zone is mapped to a floor area, allowing the system to show both a count and an estimate of people per square metre. The dashboard brings together live zone counts, recent trends and busy or congested thresholds set by the operator. Alerts flag zones that fill quickly, while dwell heatmaps show where people gather and daily reports reveal the busiest hours. Staff can use the live view to respond and the historical view to plan staffing and flow through the space. Processing runs at the edge beside the camera recorder. Frames are discarded after analysis, and the dashboard receives counts and positions without identifying the people in the scene. The result is a continuous operational view built around the cameras already installed.AIComputer visionPython~ New cameras needed0View case study
Joblogy
JoblogyJob management for trades teams.Joblogy puts every job on a single board, from enquiry to paid, with quotes, costs, photos and invoices on the job card. The schedule shows each engineer's day, unscheduled jobs ready to drag in, booked hours and travel time. In the field app, engineers see today's jobs, navigate and call the customer, track time, work through a checklist, add photos, materials and voice notes, capture a signature and take payment on site. Paid invoices sync to the accounting system so the office does not key anything twice.Web appiOSAndroidLive in productionStages from enquiry to paid6View case study
Virtual Estimator
Virtual EstimatorAn estimation portal for roofing, walling and building trades.We built an estimation portal and admin dashboard. Customers create a project, choose the project type, urgency and report type, and upload plans of up to 50 MB per file. Each order lands on a jobs board with live counts for pending, in-progress, completed, payment-pending and unpaid work. Staff move each project through a set workflow: pending, ready for modelling, ready for QA, being QA’d and completed. Projects are assigned to a modeller and every status change is recorded. Prices are set per project type and per currency, so a UK roof has a fixed price in pounds rather than a converted one. Payments run through Stripe, with weekly invoices for regular customers and a list of unpaid invoices. The admin also manages users, groups and permissions, API keys, report types, software downloads and the web store..NETBlazor WebAssemblyBootstrap~ Currencies with fixed pricing6View case study
RollBnB
RollBnBA campervan rental marketplace for Chile.We built a marketplace with search by location, dates and camper type, a map and list view of results, and filters for seats, beds, kitchen type and exterior and main equipment. Each camper has its own page with a photo gallery, category and starting price. Owners sign up with email verification and list their camper through a dedicated listing flow. A partners section groups campsites and hubs, manufacturers, and workshops and repair services, and the whole site runs in Spanish and English.WordPressListeoPHP~ Camper types on the platform7View case study
Venus Aesthetics
Venus AestheticsA clinic CRM for an aesthetics group.We built a web CRM around the clinic's day. The calendar books treatments in 15-minute slots, with each treatment colour-coded by type. The POS sells services, packages and products in one sale, applies member discounts, loyalty points and vouchers, and takes split payments by card, cash, bank transfer or client account. Behind that sit client profiles with notes and consent forms, a leads pipeline by channel, invoices and an invoice builder, a form builder, product and consumable stock across warehouses and branches, branch cash and expenses, and incentive and KPI settings for staff. The dashboard shows appointments, sales, leads and slot use live across branches.ReactTailwind CSSshadcn/ui~ Treatments in one catalogue130+View case study
Pallet Counting
Pallet CountingPallet counts from phone photos and rack occupancy from fixed cameras, reconciled with warehouse records to highlight discrepancies.We built two capture workflows around the equipment staff already use. Fixed cameras scan the aisles on a schedule, and a rack-occupancy model marks mapped positions as occupied or empty. Available locations feed the putaway list so teams can find space without a separate aisle walk. For stacks in the yard, the mobile app lets a driver or checker take a photo. The system identifies stacks, counts the visible pallets, separates pallet types and reads visible barcode labels. Staff can review the count, compare it with the delivery note and sign off the check in the same flow. Counts are reconciled with warehouse records, with mismatched locations flagged for a person to check. This turns the output of the vision system into a focused review queue, linking each observation back to the warehouse workflow instead of creating another stand-alone count.AIComputer visionPython~ Photos per stack count1View case study
Freight Dispatch
Freight DispatchDispatch and fleet operations across 14 depotsWe started with a week on the loading bay, not in a workshop, then built one scheduling model for every depot. Loads, vehicles, trailers and drivers sit in a single plan, and dispatchers drag loads onto vehicles with capacity and driver hours checked as they go. The plan is visible across depots, so loads can be moved without a phone call. Integrations bring orders in automatically, and the platform runs in the cloud so every yard works from the same live picture.Custom softwareIntegrationsCloud~ 5 monthsDispatch time−62%View case study
Invoice Reconciliation AI
Invoice Reconciliation AIInvoice reconciliation for a 900-supplier finance teamWe built a document AI pipeline that extracts lines, quantities, prices and VAT from each invoice and matches them three ways against the purchase order and goods received. The agent only posts a match it is confident about. Anything unusual, such as a price change, short delivery or duplicate, goes to an exceptions queue with the reason and the source documents side by side, so finance reviews exceptions instead of every line. Every decision is logged for audit.AI agentsDocument AIFinance~ 11 weeksMatch rate99.98%View case study
Steel Quoting Engine
Steel Quoting EngineQuoting engine for made-to-order steel fabricationWe sat with the estimator and modelled his rules for material, cutting, fabrication, finishing and margins. Then we replayed three years of real quotes through the model and tuned it until it matched his judgement. The quoting engine now builds a price from the job spec, shows the breakdown line by line, and flags anything unusual for the estimator to check. Quote turnaround went from two days to four minutes.AIPricingERP~ 9 weeksQuote turnaround2 days → 4 minView case study
Factory Floor App
Factory Floor AppOffline-first field app for nine production linesWe built the app offline-first: every form, check and photo writes to the device first and reconciles with the server when it reconnects, so the floor never waits on the network. Operators get the right checklist for their line and shift, with photos and machine readings attached. Supervisors see results as soon as devices sync, and sensor data from machines can be pulled into the same record.MobileOffline-firstIoT~ 4 monthsPaper forms removed100%View case study
Fleet Driver App
Fleet Driver AppDriver app and telematics platform for a 300-vehicle fleetWe shipped the driver app first: the day's route, stops in order, navigation, customer notes, and proof of delivery with signature and photo. Changes reach the driver instantly instead of by phone. The data from the app then fed a telematics platform for the office: live vehicle positions, stop progress, delivery exceptions and on-time performance across the fleet. On-time deliveries rose by 18%.MobileTelematicsPlatform~ 6 monthsOn-time deliveries+18%View case study
Chilled Traceability
Chilled TraceabilityTraceability from intake to dispatch for a chilled supply chainWe designed one record per pallet that follows it from goods-in to dispatch: supplier, batch, use-by date, temperature checks, location and the customer it went to. Scanning at each step keeps the record current without extra admin. A trace screen answers where a batch came from and where it went in seconds, and recall drills that took two days now take eleven minutes. We shipped module by module so the team got value early.ERPData platformCompliance~ 7 monthsUnits traced1.4MView case study
Rosettet
RosettetAn online store for a hotel toiletries brand.We built a WooCommerce store with a custom, image-led theme. Each product gets its own page with fragrance, ingredients and benefits, and a Get Quote button for trade orders. The shop has search, sorting, grid and list views, and variable products for different sizes. The home page leads with the range by fragrance, the brand's natural and vegan claims, hotel-manager testimonials and a delivery strip that states the offers plainly: free London delivery on three boxes or more, free delivery on the first order, and new 5L liquid sizes. A chat button sits on every page so buyers can ask for trade pricing without hunting for a form.WordPressWooCommercePHP~ Products in the catalogue11View all work
MedTaskly
MedTasklyA lead-gen website for medical billing.We built a custom WordPress theme with reusable page templates for services, specialties and practice sizes. That gave MedTaskly 9 service pages, 20 specialty pages, a specialties hub and 4 pages by practice size, all with the same structure and calls to action. The home page works as a sales page. It moves from the headline promise and proof points to the problem, the services grid, why MedTaskly, a four-step process, specialties, testimonials and an FAQ. A Book Your Free Audit form with a specialty dropdown sits in the hero, the audit button stays in the header on every page, and each landing page opens with a Quick Answer block written for search and AI answers.WordPressCustom themePHP~ Specialty landing pages20View all work
Pamacrafts
PamacraftsAn online store for hand-painted ceramics.We built a WooCommerce store with an image-led home page that leads with the brand's own table-setting photography, new arrivals and trending collections. Products are grouped into bowls, plates, serving dishes, spoon rests and best sellers, with sorting and grid or list views in the shop. Every product page shows a large gallery and a clear price box: piece price, delivery cost and time, total and size. A WhatsApp button sits on each product, alongside wishlist, compare and share, and a newsletter offers 10% off to bring shoppers back.WordPressWooCommercePHP~ Hand-painted pieces online39View case study
Policy & Contract Assistant
Policy & Contract AssistantAnswers from contracts, policies and procedures, with the exact clause beside each response.We built a retrieval-based assistant around the document library. Documents are split into clauses and indexed so that each question can be matched to relevant passages. The assistant uses those passages to compose a short answer, with numbered references that open the source beside it and highlight the clause. Document permissions are applied during retrieval, so answers draw only on material the person asking is allowed to access. Uploading a new document version triggers re-indexing, keeping the searchable material aligned with the library. When the available documents do not answer a question, the assistant explains the gap, shows the closest relevant material and offers a handover to the right team. An insights view brings together common questions, helpfulness feedback and unanswered topics, giving document owners a practical list of material to clarify or add.AIRAGLLM~ Every answerCitedView case study
Tender Response Drafter
Tender Response DrafterTender requirements, evidence-backed drafts and section sign-off in one bid workspace.We built a document-parsing workflow that turns the tender pack into a structured list of questions and requirements. Each scored question carries its weighting and word limit, while key dates and pass/fail checks are surfaced for the team to review early. The drafting workspace retrieves relevant material from past bids, policies, procedures and certificates. It prepares an answer within the word limit, shows its sources and marks unsupported points as requiring input. Scoring guidance appears as a checklist beside the answer, helping reviewers judge the draft against the question. Mandatory requirements are matched to evidence, with gaps assigned an owner. Named reviewers sign off individual sections before the approved material is assembled into a formatted Word or PDF response. The result is a connected bid workspace in which the team can trace a requirement through its evidence, draft and approval.AILLMRAG~ Pass/fail gapsFlagged earlyView case study
Ops Copilot
Ops CopilotPlain-English answers across job, purchasing and finance systems, with evidence and approval before action.We built a copilot that plans the steps needed to answer a question and queries the relevant connected systems. The workspace shows those steps and the supporting records beside the answer, allowing a manager to inspect how the response was assembled. Each connection has the access required for its role: read-only for looking up records and draft-only for preparing follow-up work. User permissions determine which records are available, so asking through the copilot does not widen someone’s access to the underlying systems. Where the records support a next step, the copilot prepares a draft, such as an update for a customer with a delayed job. A person reviews and approves the action before anything is sent or changed. Queries and actions are recorded in an audit log, connecting the question, evidence and follow-up in a workflow managers can review.AI agentsLLMPython~ Every queryLoggedView case study

Still running on workarounds?

FAQ

Ask AI to summarise Backendz

Still have questions?

Let’s talk — book a 15-minute intro call with our team.

Notes from the build

What we learn shipping and running production systems.

All articles