Technologies to build and improve software.
We choose technologies for new systems, portals, apps and existing solutions. Intended use, security, cost and maintenance guide each decision.
From the project objective to technical choices and a delivery that can be reviewed.
- input01
Objective and context
Idea, users, needs and constraints.
- decision02
Compared architecture
Alternatives, total cost, limits and replacement plan.
- output03
Operable system
Application, tests, documentation and operation.
Higher-impact actions require human approval.
How we choose technology
We compare alternatives against intended use and the ability to maintain the solution. Not every project needs AI, integrations or complex infrastructure.
- Use
- Objective, task, user and expected outcome.
- Data
- Origin, quality, sensitivity, retention and access.
- Integration
- Required connections, contracts and dependencies.
- Risk
- Cost of error, latency, fallback and human authority.
- Operation
- Team, support, observability and total cost.
- Evolution
- Portability, dependencies and replacement plan.
Six capabilities for your project
These capabilities support the four services. We combine only what the project needs, from a new application to system modernization.
Software, apps and integrations
We build applications, portals and apps and modernize existing systems. APIs and integrations are added when the solution needs to connect to other tools.
deliverable: Application architecture, interfaces, API contracts, tests and an evolution plan.
TypeScript · .NET · FastAPI
Criteria and application examples
- decides by
- Intended use, devices, team capability, required connections and maintenance cycle.
- boundary
- A language or framework is not introduced for novelty; it must reduce delivery or operational risk.
comparison examples
Typed application and API
Compare options for building rules and APIs for a new application or evolving existing contracts.
options considered: TypeScript · .NET · FastAPI
Portals and apps
Choose a web or mobile interface based on the customer or team task, accessibility and maintenance.
options considered: Next.js · React · Flutter
AI, agents and knowledge
Models, orchestration, retrieval and memory are assessed together because each decision changes cost, error and human authority.
deliverable: Comparative evaluation, versioned sources, tool rules, fallback and review criteria.
LangGraph · AutoGen · CrewAI
Criteria and application examples
- decides by
- Minimum quality per task, privacy, latency, authorized sources, tools and cost of error.
- boundary
- No agent expands its own scope; critical actions require an authorized tool, a record and approval.
comparison examples
Bounded orchestration
Compare how states, tools and approvals become explicit before execution.
options considered: LangGraph · AutoGen · CrewAI
Traceable knowledge
Assess retrieval, source, version and update policy for verifiable responses.
options considered: pgvector · Qdrant · LlamaIndex
Data and machine learning
Data and specialized models are considered when a baseline, target variable and measurable error exist.
deliverable: Documented dataset, reproducible experiment, metrics, limits and a monitoring routine.
scikit-learn · XGBoost · LightGBM
Criteria and application examples
- decides by
- Data quality and volume, baseline, required explainability, review frequency and cost of error.
- boundary
- A larger model does not replace suitable data, a baseline comparison or validation in the operational context.
comparison examples
Tabular baseline
Compare simpler models before justifying greater training or inference complexity.
options considered: scikit-learn · XGBoost · LightGBM
Language and representation
Assess classification, entities and similarity when the task calls for a specialized model.
options considered: BERT · Sentence-Transformers · spaCy
Cloud and infrastructure
The execution environment is chosen by volume, data sovereignty, integration and who will sustain the operation.
deliverable: Target topology, responsibilities, cost estimate, rollback and continuity plan.
Docker · Kubernetes
Criteria and application examples
- decides by
- Load, latency, data residency, observability, internal capability, total cost and reversibility of the choice.
- boundary
- Multi-cloud is not an automatic goal; dependencies and exit costs must be documented.
comparison examples
Packaging and execution
Define isolation, scale and operation without promising automatic portability.
options considered: Docker · Kubernetes
Inference under constraints
Compare hardware and runtimes by volume, latency, cost and available support.
options considered: NVIDIA · Triton · vLLM
Automation and IoT
Systems, sensors and protocols form one workflow when input, exception, decision and owner are defined.
deliverable: Signal map, protocol, frequency, exception rules, integration and contingency procedure.
MQTT · OPC UA · Modbus
Criteria and application examples
- decides by
- Physical environment, connectivity, frequency, interoperability, security and consequence of failure.
- boundary
- A critical physical action needs a safe state, local fallback and human authority proportionate to risk.
comparison examples
Industrial interoperability
Select transport and contract according to equipment, network and tolerance for unavailability.
options considered: MQTT · OPC UA · Modbus
Edge processing
Assess hardware close to operations when latency or connectivity constrain the cloud.
options considered: Raspberry Pi · NVIDIA Jetson · Arduino
Observability and governance
Telemetry, evaluation and governance connect technical behavior to an owner, a boundary and a review routine.
deliverable: Events, dashboards, alerts, evaluation criteria, runbook and a trail proportionate to risk.
OpenTelemetry · Prometheus · Grafana
Criteria and application examples
- decides by
- Criticality, required signal, response window, retention, access and evidence needed to review a decision.
- boundary
- Monitoring is not collecting everything: every signal needs a purpose, access, retention and accountable action.
comparison examples
Operational telemetry
Define traces, metrics and alerts that lead to a support or review action.
options considered: OpenTelemetry · Prometheus · Grafana
Evaluation of AI systems
Connect versions, evaluations and failures to a review routine without treating a score as final truth.
options considered: Langfuse · MLflow · W&B
Our working methods
AxionEthos and AxionCore organize engineering decisions and responsible AI use. They are internal references, applied according to the project.
- AxionEthos
- Organizes risk, explainability, human approval and records proportionate to impact.
- AxionCore
- Organizes context, components, integrations and execution boundaries for intelligent systems.
They are internal decision and implementation methods — not SaaS, a licensed platform, a certification or a separate service.
See the implementation methodComplete technology catalogue
Explore technical options by capability. Final selection depends on the project; this is not a list of mandatory tools.
View all 68 names organized by capability
Trademarks belong to their respective owners. A listed name indicates an option that may be considered; it does not claim partnership, certification, adoption in every project or automatic recommendation.
Software, apps and integrations
- Python
- TypeScript
- Rust
- Go
- Flutter
- .NET
- FastAPI
- Next.js
- React
AI, agents and knowledge
- ChatGPT
- Claude
- Gemini
- Llama
- Mistral
- Cohere
- DeepSeek
- Hugging Face
- CrewAI
- LangChain
- LangGraph
- LlamaIndex
- AutoGen
- Haystack
- MCP
- Qdrant
- Pinecone
- pgvector
- Weaviate
- Chroma
- Milvus
- Redis
Data and machine learning
- BERT
- RoBERTa
- DistilBERT
- Sentence-Transformers
- spaCy
- Transformers
- PyTorch
- TensorFlow
- scikit-learn
- XGBoost
- LightGBM
- Jupyter
- pandas
- NumPy
Cloud and infrastructure
- NVIDIA
- AWS
- GCP
- Azure
- Kubernetes
- Docker
- Triton
- vLLM
Automation and IoT
- Raspberry Pi
- Arduino
- NVIDIA Jetson
- MQTT
- AWS IoT Core
- Azure IoT Hub
- Modbus
- OPC UA
Observability and governance
- OpenTelemetry
- Grafana
- Prometheus
- Langfuse
- MLflow
- W&B
- Sentry
From technical choices to a solution in use.
Custom software, innovation, AI and automation can start with a small delivery and evolve as needs change.