FAQ
Questions we hear in discovery
Straight answers about our applied-AI studio, generative systems work and how we engage with client organizations in Canada.
Before you reach out, scan these — they cover engagement models, budgets, models and tools, data governance, and what we refuse to build.
Is Synth Forge a wellness/lifestyle brand, a music synth tool, or a deepfake service?
NO. We are an applied-AI generative studio that designs and builds generative systems and synthetic media for client organizations, always with a human in the loop. The “.life” TLD is branding only — this is NOT wellness, lifestyle or coaching. “Synth” means synthetic/generative content, NOT a music synthesizer, and we do NOT create deepfakes or deceptive impersonation. We are not a course and do not guarantee specific outputs or results.
How do you engage — project or retainer?
Most clients start with a fixed-scope generative pilot (discovery through prototype or first production slice), then move to a monthly retainer for content generation, MLOps or synthetic media pipelines. Enterprise teams sometimes run parallel workstreams — strategy roadmap plus build — under a master statement of work. We scope each path in discovery; nothing is off-the-shelf software you license from us.
What budgets should we plan for?
Indicative CAD ranges: AI strategy and discovery C$18k–C$45k; generative pilots C$45k–C$120k; full LLM copilot or content engine builds C$55k–C$165k; guardrails and evals C$28k–C$75k; MLOps and deployment C$35k–C$95k. Retainers typically start around C$7,500–C$9,500/month depending on volume and SLA. Final quotes follow a scoping call — we do not publish guaranteed ROI figures.
How long does discovery and a first prototype take?
Discovery workshops usually run two to four weeks. A generative prototype on agreed scope often lands in six to twelve weeks, depending on data pipeline readiness, integration/API complexity and review cycles. Rush timelines are possible when your team can commit reviewers daily — we will tell you honestly if the date is unrealistic.
Which models, tools and platforms do you use?
We are model-agnostic: commercial LLMs, open-weights models where appropriate, retrieval-augmented generation stacks, and your existing cloud or on-prem constraints. Tooling spans prompt engineering workflows, fine-tuning when justified, evals harnesses, and MLOps on AWS, Azure, GCP or client infrastructure. We recommend what fits your data governance — not what earns us a referral fee.
How do you handle data, IP and PIPEDA?
Client data stays under your policies unless we agree otherwise in writing. We sign NDAs and data processing terms aligned to PIPEDA. Training data, prompts and outputs ownership are defined in the SOW — typically you own bespoke prompts, fine-tunes and deliverables; we retain pre-existing tools and frameworks. Cross-border processing is disclosed upfront. Contact [email protected] for privacy requests.
What human oversight and responsible AI practices do you follow?
Human-in-the-loop review is built into interfaces, not bolted on after launch. We run model evaluation, generative guardrails, bias checks where relevant, and document limits for compliance teams. AI agents and copilots escalate low-confidence output to people. We do not promise error-free generation.
Do you monitor systems after deployment?
Yes — MLOps retainers include drift monitoring, logging, alerting and periodic eval reruns. Project-only clients receive runbooks and optional handoff workshops. Monitoring depth matches the risk of the use case; public-facing content generation gets stricter thresholds than internal draft assistants.
What are contract and notice terms?
Fixed projects use milestone billing (typically 40/40/20). Retainers bill monthly with 30-day notice to wind down. IP and confidentiality survive termination. Details are in our terms of service and your SOW.
What do you NOT do?
We do not guarantee accuracy, rankings, traffic, leads, sales, ROI or cost savings. We do not sell courses, “AI income” schemes, crypto or trading tools, MLM products, wellness apps, or deepfakes/impersonation. We are not an AGI project, not a Git forge, and not a blacksmithing shop — “forge” is our studio metaphor only.
Who owns the models and code?
Custom code, prompts, workflows and fine-tunes created for you are yours per the SOW. Our internal libraries, eval frameworks and pre-existing IP remain ours, licensed to you for the deployment. Third-party model weights and API terms follow the provider's licence — we help you stay compliant.
FAQ disclaimer: Answers describe our typical practice; your statement of work governs. Generative-AI output requires human review and can be wrong. Synth Forge does not guarantee business outcomes.