Aaziq Ali ShahMOBILE APPLICATION DEVELOPER & FOUNDERPORTFOLIO / SELECTED WORK
A closer look at the applications I’ve built, the integrations behind them, and the work currently in progress.
The mobile companion for a connected sleep device.
I built the complete app functionality: accounts, device connection, stored routines, evening and morning sequences, powernaps and playback controls. The engineering combines Flutter application state, BLE device commands and coordinated audio, light and lens tint transitions.
Read the full case study →A multimodal discovery application connecting facial recognition, description relevance and location context.
I built a Flutter application that turns lost and found reports into searchable records with images, descriptions, categories and map locations. Its matching workflow connects face detection, feature extraction, similarity scoring and ranked text retrieval to report discovery, helping people review potentially related posts and contact the person behind them.
The face processing pipeline uses MTCNN to locate faces and identify facial landmarks. Face crops are aligned and prepared for embedding extraction, reducing variation from image framing and face position before comparison.
FaceNet converts the prepared face crops into compact numerical feature vectors. Normalised embeddings are compared with cosine similarity to rank facial matches, connecting model output to candidate report retrieval rather than matching raw image pixels.
Description matching uses BM25 relevance ranking, which considers term frequency, term rarity and description length. Text scores work alongside category, date and location filters to narrow the candidate set without requiring identical descriptions.
Built authenticated report publishing, map based discovery, cloud image storage, saved posts and real time text and image conversations. Firebase Authentication, Firestore and Cloud Storage connect user identity, reports, chat rooms and media, with Google Maps supporting location selection and geocoding.
A mobile operating layer for dairy farm records, production visibility and AI assisted health insight.
I built a Flutter application for dairy farm management that centralizes animal records, milk production tracking and health related workflows. The product turns day to day farm observations into structured, searchable operational data, with an AWS backed service layer and AI model integrations that support disease risk prediction.
Designed mobile workflows to maintain cow level profiles and farm records, giving staff a practical system for reviewing animal history and operational details from the field.
Implemented structured production logging and reporting views so farms can record milk output at the animal level and compare trends over time.
Integrated AI models into the application to evaluate available animal data and surface disease risk predictions as decision support for farm operators.
Connected the Flutter client to AWS backed services for application data, remote workflows and scalable backend operations, with mobile UI states designed for dependable day to day use.
A live operating system for dispatch, drivers, deliveries and fleet visibility.
AtlasFleet was designed as a mobile-first logistics platform for teams that need to know where work is, where vehicles are, and what must happen next. It connects a dispatcher control layer with a driver execution app, turning planned routes, delivery jobs, vehicle activity and delivery evidence into one continuous operational picture.
Designed a command workflow for assigning work, prioritising jobs, monitoring delivery states and responding to operational changes without losing visibility across the day’s schedule.
Built mobile delivery workflows around job manifests, customer instructions, route context, status updates, proof of delivery, signatures, photo evidence and issue reporting from the field.
Integrated map based fleet visibility, live ETA context, route progress and delivery status so dispatchers can make informed decisions while work is still in motion.
Structured the data model around drivers, vehicles, loads, completed stops and exceptions, creating a foundation for utilisation reporting, delivery performance review and auditable service history.
A multi-location retail operating system for till operations, warehouse stock and wholesale trade.
Crazy Clearance OS was conceived as a custom in-house platform for a high-volume clearance retailer operating warehouses and stores across multiple locations. The system brings the operational layers that usually live in separate tools into one connected product: checkout, barcode-driven inventory, branch transfers, purchasing, wholesale accounts, staff activity and commercial reporting.
Designed an EPOS workflow for barcode-led sales, product lookup, discounts, overrides, refunds, exchanges, receipt printing and till access controlled by staff PINs and role permissions.
Modelled stock at product, warehouse and branch level, with SKU and barcode management, goods-in scanning, stock adjustments, damaged and returned items, movement history, low-stock signals and branch transfers.
Defined trade customer accounts, customer-specific price levels, bulk discounts, minimum-order rules, VAT handling, credit support, supplier records, purchase orders and received-versus-ordered reconciliation.
Created the basis for sales, stock value, margin, staff, location, wholesale and dead-stock reporting, alongside online-store synchronisation to keep product availability and retail or wholesale pricing aligned.
A configurable diagnostic application for accelerometer and gyroscope evaluation before gimbal PCB integration.
I built Flutter diagnostic tooling to compare sensor sampling frequencies and measurement ranges, inspect incoming device data and assess BLE delivery behaviour. The purpose was to help the hardware team identify suitable operating settings for accelerometer and gyroscope integration, with visibility into the tradeoffs between motion sensitivity, dynamic range, acquisition rate and transport throughput.
Implemented controls for sampling frequency and measurement range, including accelerometer ranges from ±2g to ±16g. The application made it possible to compare candidate configurations rather than evaluate the sensors at a single fixed setting.
Captured raw BLE notifications for hexadecimal packet inspection and timed packet counts. This exposed the device data reaching the application and supported investigation of sensor output alongside the communication path.
Compared configured sampling frequencies with notification counts over timed windows. Separating sensor acquisition from BLE delivery helped the team investigate whether increasing the sampling rate also produced the expected data flow at the application.
Worked with hardware engineers on sensor configuration for gimbal PCB integration. Adapted GATT discovery for revised UUIDs and a shared write and notify characteristic, and requested high connection priority on Android during throughput investigations.
Desktop actigraphy and sleep data analysis.
A Windows/macOS targeted application in development for tri axial accelerometer testing and sleep assessment. The planned workflow takes recordings from device acquisition through analysis and reporting.
Planned scope includes timestamped X/Y/Z storage, transfer integrity checks, sleep/wake scoring, total sleep time, latency, sleep efficiency, head position analysis and versioned re scoring. These capabilities are in development, not presented as a released or clinically validated product.
A Flutter interface for detection workflows.
Built dashboard, alert, log and report screens with HTTP/JSON integration to detection and model training endpoints. Added response parsing, request error handling, timestamped SQLite logs and chart based reporting views.
Flutter, connected devices and the interfaces between them. If that fits what your team is building, I would be glad to hear from you.