Frequently asked questions
Answers to key questions about our process, technology, and collaboration.
Process
What does a typical engagement look like?
We work in an iterative methodology: 1) Initial consultation and problem analysis, 2) Technical proposal with timelines and budget, 3) MVP development (2–4 weeks), 4) Testing and refinement, 5) Production deployment and documentation handover. Regular demos and feedback at every stage.
How long does MVP development take?
MVP timeline depends on complexity: simple projects (classification, basic NLP) — 2–4 weeks; medium complexity (object detection, chatbots) — 4–8 weeks; complex solutions (video analytics, custom models) — 8–12 weeks. We provide exact timelines for your case during estimation.
How often do we get progress updates?
Weekly calls, access to a shared task tracker (Linear/Jira), demos of new functionality as it ships. Daily short syncs for urgent projects. All materials stored in a shared workspace.
What happens after the production launch?
We hand over full documentation: API specs, deployment instructions, architecture overview. 30-day warranty for bug fixes. Additional SLA support with monitoring and model updates available.
Technical
What data do you need to start a project?
Minimum: examples of input data (text, images, video), a description of the expected output, and edge-case examples. For training we ideally want 1000+ labeled examples, but smaller datasets work via data augmentation and pre-trained models.
Can you integrate with our infrastructure?
Yes, we design around your stack. REST API and gRPC for most integrations, SDKs for Python/JavaScript. We work with AWS, GCP, Azure and on-premise. Docker and Kubernetes for deployment.
How do you ensure model performance?
Quantization, ONNX/TensorRT for inference, caching, request batching. Edge deployment for real-time tasks. Latency and throughput guarantees in the technical proposal.
How are model updates handled?
CI/CD pipeline for retraining: production model-quality monitoring, automatic triggers on metric degradation, A/B testing of new versions. Automated or scheduled updates — your choice.
Business and legal
Who owns the intellectual property?
You receive full rights to the developed solution: source code, models, documentation. Our internal libraries are licensed royalty-free. All rights transfer upon payment, contractually specified.
How is the project cost structured?
Fixed price after the spec is approved. Milestone-based payment: 30% upfront, 40% upon MVP readiness, 30% after acceptance. Long-term projects use monthly billing. Exact estimate after a free consultation.
Do you provide post-launch support?
Several tiers: basic (critical bug fixes, 30 days free), standard (48-hour SLA, security updates), extended (24-hour SLA, monitoring, model retraining). Support costs 15–25% of the project price per year.
Do you work under NDA?
Yes, we sign NDAs before discussing project details. A standard NDA covers confidentiality of data, business processes, and technical solutions.
Security and compliance
How is our data protected?
Encryption in transit (TLS 1.3) and at rest (AES-256). Isolated environments per project. Data access only for authorized staff with MFA. Regular security audits.
Are you GDPR-compliant?
Yes. We provide a Data Processing Agreement (DPA), honor the right to erasure within 72 hours, and maintain a personal-data processing register.
Can the solution be deployed on-premise?
Yes, most solutions can be deployed on your infrastructure with no external dependencies. Hardware requirements discussed per project.
Do you use our data to train other models?
No. Client data is never used to train shared models or improve services for other clients without explicit written consent. Every project is isolated; data is deleted after work completes.