Software Engineer
AI & Automation Specialist

The platform connects to CRM, e-commerce, social media, and POS systems via REST APIs and webhooks. Data is ETL'd into a unified warehouse with automated schema mapping and deduplication.

Over 200 behavioural features are computed — recency/frequency/monetary, session patterns, churn signals, seasonal trends — and fed into the ML engine for training and inference.

An ensemble of XGBoost, LightGBM, and a deep neural network generates daily predictions for churn risk, next-purchase probability, customer lifetime value, and optimal discount thresholds.

Business users explore predictions through an interactive React dashboard with drill-down filters. Slack and email alerts notify teams when a high-value customer shows churn risk above 70%.

A patient describes their symptoms via the web widget or WhatsApp. MediBot uses a fine-tuned LLM with medical context to parse the query, identify urgency, and extract key clinical entities.

The AI cross-references symptoms against a structured medical knowledge base using RAG (Retrieval-Augmented Generation) to suggest possible conditions, severity level, and recommended action — home care, clinic visit, or emergency.

If a consultation is needed, MediBot checks the clinic's real-time calendar, offers available slots, confirms the booking, and sends a calendar invite with pre-visit instructions — all within the same conversation thread.

After the appointment, MediBot sends automated medication reminders, follow-up questionnaires, and satisfaction surveys. Doctors receive a summary of the AI-patient interaction for their records.

Barcode scanners, RFID readers, and IoT weight sensors feed real-time stock data into the system. The platform reconciles physical inventory with digital records every 60 seconds, flagging discrepancies instantly.

An ML model trained on 18 months of historical data predicts optimal reorder points considering seasonality, lead times, and supplier reliability. Purchase orders are auto-generated and sent to vendors via email and SMS.

All inventory data is bi-directionally synced with Google Sheets in near-real-time, giving non-technical managers a familiar interface to monitor stock levels, edit orders, and generate reports without touching the main system.

Route planning algorithms optimise delivery schedules based on traffic, fuel costs, and order priority. Drivers receive optimised routes via a mobile PWA, and customers get live tracking links.

The app loads as a progressive web app with sub-second initial paint. All static assets are pre-cached via service workers, and product pages are pre-rendered at build time using Next.js SSG for instant navigation.

User interactions — clicks, dwell time, scroll depth, cart events — are streamed to a real-time analytics pipeline. A lightweight TensorFlow.js model runs client-side to adapt the UI based on browsing patterns.

A hybrid collaborative-filtering + content-based model generates personalised product recommendations. The model retrains every 6 hours on new interaction data, ensuring recommendations stay fresh and relevant.

Stripe-powered checkout with Apple Pay, Google Pay, and saved-card support. One-click purchase for returning customers. The entire flow is optimised for conversion, averaging 2.3 seconds from cart to confirmation.

GPS, engine telemetry, fuel sensors, and weather data from every vessel are streamed to a centralised event hub via satellite IoT. The system processes 10K+ events per second with sub-100ms latency.

A Mapbox-powered dashboard displays every vessel's real-time position, route history, ETA, and status. Operations managers can zoom from a fleet overview to individual ship telemetry with a single click.

ML models analyse engine telemetry patterns to predict component failures 48-72 hours in advance. The system automatically schedules maintenance at the nearest port, reducing unplanned downtime by 60%.

Geofencing triggers automatic alerts when vessels approach restricted zones or deviate from route. Compliance reports for port authorities and maritime regulators are auto-generated and submitted via API.