LLM Engineering RAG Systems Data Engineering MLOps AI Agents Analytics Data Science Integration Application Dev Test Services Responsible AI Prompt Engineering Data Governance Streaming Pipelines LLM Engineering RAG Systems Data Engineering MLOps AI Agents Analytics Data Science Integration Application Dev Test Services Responsible AI Prompt Engineering Data Governance Streaming Pipelines

Technical Competencies

WHAT
WE
KNOW.

12
Core technical competencies
6
Dedicated AI practice areas
50+
Technologies and platforms
0
Capabilities we claim but can't deliver
New practice — built for the frontier

AI Competencies.
Six deep.

These are new to our capability register — and they're the practice area growing fastest. We've built them through real project delivery, not classroom certification. Every competency below is backed by production systems we've shipped.

AI & Machine Learning
AI Practice

Six areas of genuine AI depth. Click any card to see full capability breakdown, tools, and what "expert" actually means for us in that area.

LLM Engineering AI · New
Large Language Model
Engineering & RAG

Production-grade systems built on top of foundation models — retrieval pipelines, prompt architecture, context management, and the evaluation frameworks that tell you if the system is actually working.

RAG Architecture
92
Prompt Engineering
95
LLM Evaluation
85
Context Optimisation
88
Multi-stage RAG with hybrid search (dense + sparse retrieval)
Chunking strategy, embedding model selection, and index optimisation
Prompt chaining, tool use, and structured output extraction
Hallucination detection and citation-grounded response generation
LLM-as-judge evaluation pipelines with automated regression testing
Multi-model routing — directing queries to cheapest/fastest capable model
Enterprise access control mapped to retrieval permissions
Latency optimisation, caching strategies, and cost management at scale
LangChainLlamaIndexWeaviatePineconeQdrantOpenAI APIAnthropic APIRAGASArize Phoenix
Expand+
Agent Systems AI · New
AI Agent Development
& Orchestration

Multi-step, multi-tool AI agents that operate inside real business workflows. We engineer the guardrails, escalation logic, and audit trails that make agents safe to deploy in production — not just impressive in demos.

Agent Architecture
88
Tool Use & APIs
92
Safety & Guardrails
90
Multi-Agent Systems
80
ReAct and Plan-and-Execute agent patterns for complex multi-step tasks
Tool and function calling across enterprise APIs, databases, and services
Human-in-the-loop escalation gates for ambiguous or high-risk decisions
Full audit logging of agent actions, reasoning chains, and tool calls
Multi-agent coordination — supervisor/worker and peer-to-peer patterns
State management and long-running agent session handling
Failure recovery, retry logic, and graceful degradation patterns
Cost and latency budgeting per agent task to prevent runaway spend
LangGraphAnthropic ClaudeAutoGenCrewAIClaude CodeOpenAI AssistantsTemporalMCP
Expand+
Model Customisation Specialist
Fine-tuning &
Model Adaptation

When prompting isn't enough. We fine-tune open-weight models on proprietary data, adapt foundation models to specialist domains, and run the evaluations that prove the adapted model is actually better — not just different.

Supervised Fine-tuning
85
LoRA / QLoRA
88
Dataset Curation
90
Eval & Benchmarking
82
Parameter-efficient fine-tuning with LoRA and QLoRA on consumer and cloud hardware
Instruction tuning dataset construction from existing business content
Domain adaptation for legal, medical, financial, and technical language
RLHF-lite approaches using preference data from subject matter experts
Custom evaluation harnesses benchmarked against domain-specific tasks
Private / on-premise deployment using Llama, Mistral, Phi families
Llama 3.3MistralHugging FaceAxolotlUnslothWeights & BiasesPEFTvLLM
Expand+
AI Infrastructure Specialist
MLOps &
AI Platform Engineering

The operational layer that keeps models working after launch day. Feature stores, model registries, drift detection, automated retraining — because a model that worked at go-live and fails six months later is worse than not shipping one.

Model Registry
90
Feature Stores
85
Drift Monitoring
88
Retraining Pipelines
82
Experiment tracking, model versioning, and lineage across training runs
Feature store design for training/serving consistency and preventing leakage
CI/CD pipelines for automated model training, evaluation, and promotion
Data drift, concept drift, and model performance monitoring with alerting
Shadow mode and A/B testing frameworks for safe model rollouts
Model serving infrastructure with auto-scaling and latency SLAs
MLflowWeights & BiasesFeastVertex AI PipelinesSageMakerArize AIWhyLabsBentoML
Expand+
AI Governance Specialist
Responsible AI &
Safety Engineering

Enterprise AI that passes legal, risk, and compliance scrutiny. We build the explainability layers, bias audits, and governance frameworks that turn AI from a liability question into a documented, auditable, defensible system.

Explainability
88
Bias & Fairness
85
AI Risk Frameworks
90
Regulatory Alignment
80
Model card creation and capability/limitation documentation
SHAP and LIME explainability for ML model decisions
Bias detection across protected characteristics in model outputs
AI risk tiering aligned to EU AI Act and ISO 42001 frameworks
Prompt injection and jailbreak hardening for LLM-powered applications
Human oversight design — escalation triggers and decision audit trails
SHAPLIMEFairlearnAI Fairness 360Guardrails AILLM GuardISO 42001
Expand+
Product Engineering AI · New
Generative AI
Product Development

End-to-end engineering of user-facing AI products — chat interfaces, document intelligence tools, AI-augmented workflows. We handle the full stack: UX, AI backend, integration layer, safety layer, and the observability to know if it's working.

AI UX Patterns
88
Streaming Responses
92
Cost Optimisation
85
Multi-modal
80
Conversational UI design with streaming, citations, and structured outputs
Document intelligence — extract, classify, and reason over PDFs and images
Multi-modal pipelines combining vision, text, and structured data
Feedback loops and thumbs-up/down collection for continuous improvement
Token cost tracking, caching strategies, and prompt compression
A/B testing frameworks for model versions and prompt variants
Next.jsVercel AI SDKAnthropic APIOpenAI APIHeliconeLangfuseBraintrust
Expand+
Data Practice
Data Engineering,
Science & Analytics

The original three — now with AI deeply embedded at every layer. These aren't unchanged from 2019; they've been rebuilt around AI-era data requirements.

Foundation AI-Enhanced
Data Engineering

Scalable, observable, AI-ready data pipelines. We build the infrastructure that makes every downstream use case — analytics, ML, AI agents — reliable enough to trust.

Pipeline Architecture
95
Stream Processing
88
Data Quality
92
Cloud Platforms
90
Lakehouse architecture on Databricks, Snowflake, and BigQuery
Medallion architecture (Bronze/Silver/Gold) with data contracts
Real-time streaming with Kafka, Kinesis, and Flink
dbt transformation layer with CI/CD and automated testing
AI-assisted pipeline generation and documentation (new)
End-to-end data lineage and observability with Monte Carlo
Vector store integration for AI-ready data products
ERP and CRM source system integration (SAP, Salesforce, Workday)
DatabricksSnowflakedbtAirflowKafkaSparkTerraformGreat ExpectationsAtlan
Expand+
Modelling AI-Enhanced
Data Science &
Machine Learning

Classical ML and statistical modelling — with LLMs used to accelerate the research, feature engineering, and evaluation phases. We choose the right model for the problem, not the most impressive one.

Classical ML
93
Deep Learning
85
Time Series
90
Computer Vision
82
Demand forecasting, churn prediction, anomaly detection at scale
Gradient boosting (XGBoost, LightGBM) for tabular data problems
Time series with Prophet, NHiTS, and Temporal Fusion Transformers
Computer vision for inspection, surveillance, and document processing
LLM-assisted feature engineering and EDA (new)
Causal inference and uplift modelling for intervention design
scikit-learnPyTorchXGBoostProphetYOLOPandasPolarsMLflow
Expand+
Intelligence AI-Enhanced
Data Analytics &
Business Intelligence

Analytics redesigned for the AI era — NL query interfaces, semantic layers, and decision-driven design. We build analytics that changes behaviour, not dashboards that collect login attempts.

Semantic Layer
92
BI Platforms
95
NL Analytics (AI)
85
Self-serve Design
90
Tableau, Power BI, Looker implementation with governed semantic layer
dbt semantic layer and metric governance — one definition, everywhere
LLM-powered natural language query interface over data warehouse (new)
Decision-first analytics design — built backwards from the decisions
Row-level security, certified datasets, and embedded analytics
AI-generated narrative insights and anomaly commentary (new)
TableauPower BILookerMetabasedbt SemanticCube.devMonte Carlo
Expand+
Engineering Practice
Application, Integration
& Test Engineering

The engineering backbone. AI accelerates delivery here — but the craft of building reliable, maintainable systems at enterprise scale is still a human discipline.

Development AI-Assisted
Application
Development

Full-stack enterprise application development — with AI co-pilots embedded across the engineering lifecycle. We ship faster without cutting the corners that cause production incidents.

Full-stack Web
93
API Design
92
Mobile
80
AI-Embedded Apps
90
React, Next.js, TypeScript frontend engineering
Python (FastAPI, Django), Node.js, Go backend services
AI-embedded internal tools and workflow automation
Enterprise SSO, RBAC, and compliance-ready architecture
AI co-pilot assisted development with Cursor and Claude Code
Performance engineering, CDN optimisation, and load testing
ReactNext.jsFastAPIPostgreSQLRedisKubernetesCursorClaude Code
Expand+
Connectivity AI-Enhanced
Integration
Services

Making systems talk to each other reliably — and making AI capabilities connect to enterprise systems without breaking them. Event-driven, resilient, and always observable.

API Platform
92
Event Streaming
88
ERP/CRM Connect.
85
AI System Connect.
90
REST, GraphQL, and gRPC API design and gateway management
Event-driven architecture with Kafka, RabbitMQ, and AWS EventBridge
SAP, Salesforce, and Workday integration patterns
AI service integration — connecting LLMs to enterprise workflows safely
Dead-letter queues, circuit breakers, and idempotent retry logic
MCP (Model Context Protocol) server implementation for AI tool use
KafkaMuleSoftAWS EventBridgeKong GatewayBoomiMCPPostman
Expand+
Quality AI-Accelerated
Test & Quality
Assurance

AI-accelerated test engineering — from unit to E2E, plus the AI-specific quality concerns that standard QA frameworks don't cover: LLM output validation, prompt regression, and agent behaviour testing.

Test Automation
92
Performance Testing
88
AI Output Testing
85
Security Testing
80
AI-generated unit and integration test suites with Codium
Playwright and Cypress E2E test automation
LLM output validation and hallucination regression testing
Prompt injection and jailbreak penetration testing for AI apps
Load and performance testing with k6 and Gatling
Shift-left QA embedded in CI/CD pipeline from sprint one
PlaywrightCypressCodiumk6RAGASPyTestSnykOWASP ZAP
Expand+
Depth at a glance

Capability register

Depth rating out of 5. Assessed against real project complexity, not certification status. Certifications shown where held.

Competency Depth AI Layer Certifications / Standards Years Active
─── AI & Machine Learning
LLM Engineering & RAG
Core practice 3 yrs
AI Agent Development
Core practice 2 yrs
Fine-tuning & Model Adaptation
Core practice 2 yrs
MLOps & AI Platform
Core practice AWS ML Specialty 4 yrs
Responsible AI & Governance
Core practice ISO 42001 2 yrs
Generative AI Product Dev
Core practice 2 yrs
─── Data Practice
Data Engineering & Pipelines
AI-enhanced Databricks Associatedbt Certified 8 yrs
Data Science & ML
AI-enhanced GCP Professional DS 7 yrs
Analytics & BI
AI-enhanced Tableau Desktop Cert 9 yrs
─── Engineering Practice
Application Development
AI-assisted AWS Solutions Arch. 10 yrs
Integration Services
AI-enhanced MuleSoft Developer 8 yrs
Test & Quality Assurance
AI-accelerated ISTQB Advanced 9 yrs
Technology ecosystem

Tools we work with
every day.

Primary — used on most projects   Secondary — used where appropriate   Available — certified but specialist

AI Platforms
Anthropic Claude Primary
OpenAI GPT-4o Primary
xAI Grok 3 Secondary
Google Gemini 2.0 Secondary
Meta Llama 3.3 Secondary
Mistral Large Secondary
AWS Bedrock Primary
Azure OpenAI Primary
Data Platform
Databricks Primary
Snowflake Primary
BigQuery Primary
dbt Core & Cloud Primary
Apache Airflow Primary
Apache Kafka Primary
Apache Spark Primary
Monte Carlo Secondary
AI Engineering
LangChain / LangGraph Primary
LlamaIndex Primary
Weaviate / Pinecone Primary
MLflow Primary
Weights & Biases Primary
Cursor IDE Primary
Claude Code Primary
Hugging Face Secondary
Engineering Stack
React / Next.js Primary
Python / FastAPI Primary
Node.js Primary
Kubernetes Primary
Terraform Primary
GitHub Actions Primary
Playwright Primary
Kafka / RabbitMQ Primary
01
We use these tools ourselves

Every competency listed is used on live projects, not just in certifications. Our own delivery process runs on the same AI tools we recommend to clients. We know where they break because we've experienced it.

02
We say what we can't do

If a capability is out of scope for us, we tell you — and recommend someone who can. We won't staff a project with people who are learning the stack on your budget. Depth ratings are honest, not aspirational.

03
AI competencies are current

The AI practice is the fastest-moving space in engineering. We run internal model evaluations monthly, rotate primary tools as capabilities shift, and train the team continuously. The competency register is reviewed quarterly.

Want to see
these in action?

Tell us the problem you're trying to solve. We'll map it to the right combination of competencies — and be straight with you about where we can genuinely add value and where you should look elsewhere.

Talk to us