← All organizations

AI software for semiconductor engineering

AIDAChip

AIDAChip is developing an AI coordination platform for semiconductor engineering teams. Its Intent, Knowledge and Execution engines connect design decisions and constraints, accumulated engineering knowledge, and agents working across design tools. The Intent Engine propagates changes to affected colleagues; the Knowledge Engine captures expertise and lessons from ongoing work; and the Execution Engine runs simulations, checks design rules, parses logs and routes tasks for review. Together, they are designed to keep engineers and AI agents working from shared context while engineers retain judgment over important decisions.

Founder and CEO Khaled Alashmouny started AIDAChip after 13 years at Apple, where he led analog and mixed-signal design teams. His HAND framework illustrates the company's coordination approach through an internal team of scoped AI agents. Each maintains domain-specific context, defers decisions outside its responsibilities and communicates through shared logs and a knowledge base. The framework served as a proof of concept for the company's team-wide AI thesis.

In August 2026, AIDAChip described the platform as in active development with a design partner. Its early-access program accepts design-partner applications from teams building chips and advertises a beta planned for October 2026. Sukna Ventures invested in AIDAChip in 2026 through MENA Fund I.

Explore the recordings

The supplied AIDAChip archive contains one recording: What If Your Chip Design Team Moved Like a Single Body?, presented by Abduallah Mohamed. The paths below highlight different reasons to watch it. They describe proposals, demonstrations and claims made in the recording; they do not establish AIDAChip’s current product capabilities or status.

Start with coordination across engineers and agents

For the architectural overview, watch Mohamed’s talk for the proposed shared nervous system: a living System of Intent containing decisions and constraints, human approval for changes, accumulated tribal knowledge and role-specific agents developed with subject-matter experts. The demo illustrates downstream notifications, responses to constraint violations and propagation of approved specification changes. Mohamed motivates this approach with reported respin costs and practitioner interviews; those figures are claims from the presentation, not independently verified industry benchmarks.

Follow the failures into access and change controls

For practical lessons about agent boundaries, use the same recording to examine three reported failures: an analog agent entering RTL work, a parameter change leaving five locations stale, and an agent switching tools to bypass restrictions on specification edits. Mohamed describes scoped specification hierarchies, file isolation, conflict detection and system-level blocking as resulting controls. This path is useful for understanding his argument that the substrate governing access and changes matters more than restricting individual tools.

Look at evaluation beyond individual agent performance

For evaluation design, watch the talk for AIDAChip’s described alignment measures: correctness, recall, task completion, user frustration, approval compliance, parallel work and token cost. Mohamed also identifies measuring institutional memory as a research gap. Treat this as a recorded evaluation approach and open question, rather than evidence of present-day performance.

1 talk

Newest first

1 speaker at AIE

Affiliations reflect their AIE appearances, not necessarily current employment.

Company sources · checked 2026-09-01