Technology
@I Design links brief interpretation, component selection, schematic generation and layout optimisation through one shared design representation, with a rule engine governing every AI proposal.
Design intent is carried forward explicitly. Each requirement keeps a traceable link to the components, nets and layout decisions that implement it, and reviewed outcomes feed back into how future briefs are interpreted.
Core Modules
Six modules, one shared data model, and a rule engine with the final say.
No module works in isolation. A change at any level (requirement, circuit or layout) propagates consistently to the others, so the schematic, the board and the intent behind them stay aligned.
Module 01
Turns a written brief into structured engineering intent.
Extracts voltages, interfaces, environmental limits and implied requirements, scores its confidence in each, and flags uncertain items for review instead of guessing.
Module 02
Matches parts to the requirement, not to habit.
Ranks candidates on electrical fit, cost, lead time and reliability data, and records the reason each part was chosen.
Module 03
Builds netlists and schematics from proven circuit patterns.
Parameterised patterns for power rails, drivers, protection and filtering are combined into a complete schematic in standard EDA formats.
Module 04
Places and routes against competing objectives.
Searches for Pareto-optimal layouts across trace length, via count, EMI exposure, thermal hotspots and signal integrity, so trade-offs are shown rather than hidden.
Module 05
Enforces what must not be negotiated.
Electrical safety, manufacturing and design rules are checked before, during and after generation. Safety-critical rules hold a veto over optimisation.
Module 06
Improves from reviewed outcomes.
Where an engagement permits it, reviewer edits, DRC results and acceptance decisions refine interpretation and optimisation through versioned, regression-tested model releases.
Governance Boundary
Adaptive inference works inside hard limits. Safety, regulatory and manufacturing constraints are enforced by deterministic rules, not learned.
Decision domains are separated explicitly. The AI handles component preferences, layout objectives and design-pattern choices; the rule engine owns safety, regulatory requirements and manufacturing limits. When they conflict, a staged fallback applies: relax within safe margins, try an alternative approach, fall back to a proven pattern, or stop and hand the issue to an engineer with specific guidance.
Integration
Outputs arrive in standard formats, so your engineers review and finish the design in the tools they already use.
A tool-abstraction layer connects the pipeline to external EDA tools. A typical release package contains KiCad-compatible schematic and board files, a structured BOM, a DRC report and industry-standard Gerber files. Exact formats are confirmed at intake.