Case study · Healthcare
AI-initiated medication adherence calls in home care
A deployment study of voice agents used to check patient medication compliance — protocol design, failure modes, and caregiver response patterns.
Field report · Technology
A streaming AI avatar as competitive judge
Findings from deploying a real-time generative avatar to deliver hackathon verdicts — participant reception, latency constraints, and the effect on feedback uptake.
Case study · Education
Community-owned speech archives with dual transcription models
A tool built for minority language communities to record, transcribe, and preserve speech — comparing Whisper and wav2vec2 accuracy across low-resource language samples.
Case study · Engineering
Digital twin and discrete-event simulation for packing line optimisation
Deploying a 3D digital twin alongside a SimPy simulation engine on a commercial cucumber packing line — throughput gains, staffing decisions, and bottleneck identification.
Field report · Architecture
A five-stage AI pipeline from product photograph to 3D spatial asset
Documenting a full content pipeline: background removal, image-to-mesh generation, environment placement, and Gaussian Splat integration — with failure analysis at each stage.
Playbook · Science & Research
Applying the Bradley-Terry model to AI-assisted competitive ranking
Using pairwise comparison and a statistical ranking algorithm to evaluate creative and technical submissions — a methodology for reducing scoring bias in structured competitions.
Field report · Technology
AI agents as first-class architectural components in game design
A development study of embedding AI agent logic directly into game systems rather than as a post-hoc feature — design patterns, decision surfaces, and emergent behaviour.
Case study · Agriculture
Evidence-traced crop planning using retrieval-augmented generation
Building a RAG system for Western Canadian dryland operations in which every AI-generated recommendation is linked to a source document — methodology, agronomist trust metrics, and retrieval accuracy.
Playbook · Agriculture
A domain rules layer between field data and the language model
How encoding agronomic validation logic upstream of an LLM reduces hallucination in crop plans — rule design, failure cases, and the effect on agronomist correction rates.
Case study · Healthcare
AI-generated shift handoff notes in home care settings
A longitudinal study of caregiver-to-caregiver handoff summaries produced by a language model — information fidelity, correction behaviour, and care continuity outcomes.
Case study · Healthcare
Risk stratification from longitudinal care logs
Building a patient risk model from daily care and health log data — signal selection, threshold calibration, and comparison with clinician-flagged cases over a six-month period.
Field report · Engineering
LLM-assisted security auditing across six vulnerability classes
A structured evaluation of AI code review applied to authentication, dependencies, environment variables, secrets, and row-level security — benchmarked against manual audit findings.
Field report · Science & Research
Whisper and wav2vec2 in low-resource community language transcription
A comparative study of two transcription architectures applied to non-dominant language recordings — accuracy divergence, confidence scoring, and implications for community-controlled linguistic archives.
Case study · Agriculture
Operational performance of a domain-narrowed GPT in commercial horticulture
A field evaluation of a custom GPT configured with deep operational context for a single greenhouse — comparing output quality and staff adoption against a general-purpose baseline.
Playbook · Science & Research
A six-pass sequential prompting framework for evaluating technical submissions
A structured multi-pass evaluation architecture in which each prompt targets a distinct analytical dimension — design rationale, prompt construction, and inter-pass consistency findings.
Playbook · Business
Webhook-driven data routing for multi-agent voice deployments
A lightweight architecture for logging AI call outcomes across multiple voice agents into separate structured records — applicable to businesses without a CRM layer.
Case study · Business
Automated outbound AI calling for small business customer acquisition
A deployment of rate-limited batched outbound voice calls with real-time webhook tracking — call volume, outcome distribution, and the threshold at which human follow-up became necessary.
Case study · Business
Lead scoring trained on AI voice call summaries
Using call outcome data from a voice agent as the primary signal for lead qualification — model construction, comparison with form-based scoring, and precision on booked conversions.
Playbook · Business
Live calendar integration as a closing tool in AI voice calls
A tool integration pattern enabling a voice agent to check availability and book appointments in real time — implementation architecture, fallback design, and booking completion rates.
Case study · Business
Voice analytics as a CRM intelligence layer in automotive retail
Building structured dashboards from accumulated AI call summaries in a tire shop CRM — what call data surfaces about customer intent and vehicle history that form-based records do not.
Case study · Business
Semantic matching for AI service onboarding using document embeddings
An OpenAI embedding system that reads a prospect's company documentation and matches it to relevant service offerings — compared against intake form outputs across 40 onboarding sessions.
Playbook · Business
Scheduled AI call synthesis as a team intelligence workflow
An automated weekly pipeline that categorises call summaries, generates a visual breakdown, and distributes a formatted digest — and the operational case for scheduled synthesis over live dashboards in small teams.