Home About

Solutions

AI & Agentic Automation Business Analytics Data Engineering Dashboards & Visualization Technical Training

Products

In-Sight Analytics In-Sight Vision GigaSave Fola Case Studies Book a Discovery Call

Real Problems.
Measurable Results.

We don't deal in demos and slide decks. Here is the work we have done, the challenges our clients faced, and the numbers that prove it.

Mining · Computer Vision Richards Bay Minerals

Eliminating Calcrete Blockages Before They Stop Production

The Challenge

Richards Bay Minerals, a world-leading heavy mineral sands producer, faced a recurring and costly problem: calcrete blockages forming in their pond pump operations. These blockages were invisible until they caused full pipeline blockage events, each capable of shutting down operations for up to 4 days at a time, at enormous cost to production and maintenance schedules.

Detection was entirely manual and operator-dependent. The risk of a missed event was always present. The team needed a system that could watch the ponds continuously, and respond faster than any human could.

The Solution

Excite-Data deployed an In-Sight Vision computer vision system tailored to the specific visual signature of calcrete formation. Over 40,000 training images were captured and annotated across varying lighting, weather, and operational conditions at the site. A deep learning detection model was trained and optimised for edge deployment directly at the pond.

The system integrates with the existing PLC infrastructure, triggering immediate alarms and automated responses the moment calcrete is detected, with a response time under 1 second.

Key Technologies Used

  • Deep learning object detection (convolutional neural network)
  • Edge compute deployment for sub-second inference
  • PLC alarm integration for automated plant response
  • 40,000+ site-specific training images for high accuracy
  • 24/7 continuous monitoring with no operator dependency

Results

<1s
Detection response time
24/7
Continuous automated monitoring
40K+
Training images collected & annotated
4-day
Blockage risk detected before downtime

Want a similar solution for your operation?

Talk to Us

Manufacturing · AI Vision Aluminium Hot Rolling Mill

R11M Annual Recovery via Automated Crop Eye Detection

The Challenge

At a high-throughput aluminium hot rolling facility, the cropping of ingot heads and tails was performed manually by operators making real-time judgement calls. The process was inconsistent, operators working under pressure would leave excessive material margins to be safe, resulting in significant over-cropping on every single pass.

With the volume of ingots processed per day, even a conservative estimate of excess material loss per pass compounded into millions of rands in annual material write-off. The team needed a deterministic, vision-based system that could locate the crop eye exactly, and trigger the cut at precisely the right point, every time.

The Solution

Excite-Data built and deployed an automated crop eye detection system: a YOLOv4 object-detection model trained on labelled production imagery identifies the crop eye position in real time from dual live camera feeds on the rolling line. Detections stream to the mill control system over a ZeroMQ interface, providing automated cut-point recommendations with PLC-ready triggers for the Heavy Shear.

The process context: 630 mm ingots are reduced to roughly 18 mm over 30 consecutive passes, and every over-cautious manual cut compounds. Cutting at the detected point recovers up to 50 kg of material per side per pass, worth about R30,180 per day at production volumes.

Key Technologies Used

  • YOLOv4 detection model trained on labelled production imagery
  • Real-time camera integration on the rolling line
  • Automated cut trigger via mill control system integration
  • Operator dashboard for monitoring and override

Results

R11M
Projected annual material savings
50kg
Material recovered per side per pass
R30k
Material value recovered per day
24/7
Consistent cuts, every pass, every shift

Want a similar solution for your operation?

Talk to Us

Mining · Data Quality · Predictive Maintenance Royal Bafokeng Platinum · Styldrift

From Data Quality to Asset Intelligence at Royal Bafokeng Platinum's Styldrift Mine

The Challenge

Royal Bafokeng Platinum's Styldrift Mine runs a WiFi-enabled underground smart mining platform generating continuous sensor data from its mobile fleet: face drill rigs, roof bolters, LHD vehicles, and utility vehicles. Before that data could be trusted for analytics or predictive maintenance, one question had to be answered honestly: is it any good?

At the same time, the reactive maintenance model was costly: high callout fees, long parts lead times, and unplanned stoppages eroding margins month over month.

The Foundation: Data Quality First

Excite-Data began where trustworthy analytics has to begin. We built a custom ETL pipeline from the mine's Azure data store and ran a formal data quality study across more than 20 million records and 190+ variables from the four machine classes: completeness, duplicates, statistical integrity, anomaly rates, and fitness for machine learning. The study surfaced the silent killers of mining analytics, including sensors flatlining at constant values across millions of readings, and delivered a governance report the client could act on.

The Solution

On that verified foundation, Excite-Data deployed In-Sight Analytics for the four critical asset classes: live sensor ingestion, ML models trained on historical failure signatures, real-time health scores with early warning alerts, and SPC charts on a custom operations dashboard, giving maintenance teams the visibility to intervene during planned windows instead of scrambling during unplanned ones.

Styldrift was the first production home of In-Sight Analytics, today a full plant-intelligence platform at insight-analytics.io.

Key Deliverables

  • Formal data quality study: 20M+ records across four machine classes
  • Custom ETL pipeline from Azure blob storage; data governance report
  • ML-based failure prediction models (95%+ accuracy)
  • Real-time health scoring, SPC charts, and anomaly detection
  • Custom operations dashboard for maintenance teams

Results

95%+
Failure prediction accuracy
20M+
Records quality-checked before a single model was trained
4
Critical asset classes monitored in real time
1st
Production home of In-Sight Analytics

Want a similar solution for your operation?

Talk to Us

Energy · Analytics Platform Candi Solar

Tariff & Savings Intelligence for a Solar-as-a-Service Provider

The Challenge

Candi Solar provides solar-as-a-service to commercial and industrial clients across South Africa. As the fleet grew, three problems compounded: tariff benchmarking was inconsistent across clients, ROI calculations were opaque to customers, and there was no real-time view of generation versus consumption. Reporting was manual and reactive, consuming engineering time that should have gone into the fleet.

The Solution

Excite-Data built a real-time energy intelligence platform: live ingestion from solar inverters and IoT meters over MQTT, a Python processing layer computing solar yield, consumption, and lifetime CO₂ savings into a time-series database, and a Savings & Tariff Intelligence dashboard giving each customer transparent Candi Tariff, Lifetime Tariff, and ROI views with drill-down.

The platform runs on a multi-tenant cloud architecture with role-based access governance aligned to POPIA and ISO 50001, and automates customer segmentation and ESG reporting, including carbon accounting.

Key Technologies Used

  • Next.js + Django REST Framework + PostgreSQL
  • MQTT ingestion from inverters and IoT meters
  • Time-series analytics: yield, consumption, lifetime CO₂
  • Multi-tenant RBAC aligned to POPIA and ISO 50001

Results

R1.8M
Annual billing corrections identified
20%
Tariff accuracy improvement
65%
Reduction in per-client engagement time

Running a distributed energy fleet?

Talk to Us

Facilities · Engineering Consulting Empact Group · Siemens Midrand Campus

Data-Driven HVAC Assessment for the Siemens Midrand Campus

The Challenge

Empact Group manages the facilities of Siemens' office complex in Midrand: six office buildings with tenants reporting thermal discomfort attributed to the ventilation system. Empact needed more than opinions; it needed an evidence-based answer on whether the HVAC estate was performing, compliant, and correctly sized.

The Solution

Excite-Data ran a two-day site assessment across all six buildings, then let the data speak. Interval electricity metering across the campus zones was analysed for consumption patterns and assessed against the SANS 10400-T energy efficiency standard. In parallel, we built an ASHRAE-based capacity model of the complex: required airflow rates per space type, and the matching Fan Air Terminal, Air Handling Unit, and chiller capacities for occupancy of up to 140 people per 700 m² floor, including optimal fresh-to-return air ratios under different ambient conditions.

The result was a set of engineering recommendations Empact could action and defend: what was compliant, what was undersized, and what settings would hold comfort without wasting energy.

Key Deliverables

  • Six-building site assessment and HVAC system evaluation
  • Interval-metering consumption analysis, SANS 10400-T compliance
  • ASHRAE-based ventilation and cooling capacity model (FAT, AHU, chiller)
  • Ventilation best-practice recommendations report

Results

32%
Lower monthly consumption across assessed months (16,546 → 11,288 kWh)
6
Office buildings assessed end to end
SANS
10400-T compliance verified with evidence

Need an evidence-based answer on your facilities?

Talk to Us

Retail · Technical Training Woolworths

Training the First Cohort of the Woolworths IT Engineering Academy

The Challenge

Woolworths launched its IT Engineering Academy to grow its own engineering talent, and the founding cohort of graduates needed to make the jump from academic knowledge to production-ready skills. That takes instructors who build systems for a living, not career trainers reading from someone else's slides.

The Solution

Excite-Data designed and delivered the cohort's technical curriculum across four tracks: Data Science, Software Engineering, Data Engineering, and workplace readiness. Training ran as hands-on, instructor-led workshops in which every graduate built working systems: ETL pipelines processing multi-format data, REST APIs on Django, and relational schemas across MySQL and PostgreSQL, all managed with Git and delivered Scrum-style, the way real teams work.

The workplace-readiness thread ran through everything: version control discipline, code review, agile ceremonies, and the habits that make a graduate useful in their first sprint, not their first year.

Curriculum Delivered

  • Python for Data Science, Machine Learning, and Web Development
  • Django REST Framework: build and ship real APIs
  • SQL foundations plus MySQL and PostgreSQL in practice
  • Git version control and Agile delivery with Scrum

Results

1st
Founding Academy cohort trained end to end
9
Hands-on modules across four tracks
100%
Project-based: every graduate shipped working code

Building in-house tech capability?

Talk to Us

Public Sector · Data Governance National Social-Protection Agency

A Data Quality & Governance Strategy for 19 Million Beneficiaries

The Challenge

A national agency disbursing social grants to 19 million beneficiaries faced a data problem with fiscal consequences: in one year, roughly seventy-five thousand payments worth about R140 million went to deceased beneficiaries because death notifications arrived too late in the payment cycle. Beneficiary records were fragmented across legacy systems, and data sharing with other government departments ran on manual, batch-based exchanges.

The Solution

Excite-Data delivered a data quality and governance strategy for the agency's leadership: an enterprise governance operating model built on DAMA-DMBOK functions with ISO 8000 quality dimensions and a POPIA controls matrix; a Master Data Management blueprint consolidating legacy systems into a single beneficiary registry keyed on the national ID; and an interoperability architecture for real-time verification against authoritative sources, replacing monthly batch files with API and event-driven exchange.

The plan is phased over 36 months, sequenced to bank quick wins first: deduplication and real-time validation halt payments to deceased beneficiaries in the opening months, while the registry, interoperability fabric, and AI-assisted anomaly detection build the long-term capability. A balanced scorecard defines the KPIs, baselines, and reporting cadence for the whole programme.

Key Deliverables

  • Data governance operating model: DAMA-DMBOK, ISO 8000, POPIA controls
  • Master beneficiary registry blueprint (MDM, survivorship rules)
  • Real-time interoperability architecture with authoritative registries
  • Phased 36-month implementation plan with balanced scorecard

Scope & Targets

19M
Beneficiaries served by the systems in scope
R140M
Annual leakage the plan's first phase targets
36
Month phased roadmap, quick wins first

Wrestling with data quality at institutional scale?

Talk to Us

Transport · Data & Analytics Bluebird Group

From Data Review to an Intelligence Dashboard for Bluebird Group

The Challenge

Bluebird Group runs staff transport across Cape Town, Johannesburg, Durban, and George, with hundreds of thousands of trips flowing through a production database that had grown organically over years. Before leadership could trust any dashboard built on it, a harder question needed answering: what can this data actually tell us, and where does it fall silent?

The Foundation: Data Review First

Excite-Data began with a structured review of the production snapshot: schema topology, logical data marts, table-by-table freshness, and a quantified findings register covering compliance capture rates, dormant lookup tables, and values that could not be recovered from the database at all. Every finding was worked through with Bluebird's engineers in a structured question-and-answer cycle, so the intelligence layer would be built on verified ground rather than assumptions.

The Solution

On that foundation, Excite-Data built the Bluebird Intelligence Dashboard: a web application querying the live operational database directly, with five executive views covering the whole business: Executive Summary, Operations Intelligence, Fleet & Maintenance, Drivers & Workforce, and Financial Intelligence. Every KPI card, chart, and table drills down to the underlying records, and the schema-introspecting architecture required no migrations and no ETL layer on Bluebird's systems.

Key Technologies Used

  • Structured data review with a quantified findings register
  • Next.js server components querying MySQL directly
  • Type-safe schema introspection: zero migrations on the client's database
  • Interactive ECharts visualisations with record-level drill-down

Results

5
Executive intelligence views, one operational picture
4
Cities covered in a single dashboard
100%
Of charts drill down to the underlying records

Running operations on a database you don't fully trust?

Talk to Us

Your Operation Could Be
Our Next Case Study.

Book a free discovery call and let us assess where your greatest data opportunities lie. No obligation, just a straight conversation about what's possible.