Stories · Manufacturing
40% more throughput on the clusters that ship chips : delivered the agent-native way.
Synaptics' EDA cluster bottleneck was a delivery economics problem as much as an engineering one. A traditional SI would have proposed quarters of analyst hours mapping legacy infrastructure by hand. Mactores ran the same discovery in days using Aedeon : and reallocated the engagement budget into the smart-queue architecture work that actually moved the throughput numbers.
Baseline
Weeks of queue contention on legacy infrastructure. Throughput drag on chip-design cycles.
Outcome
40% throughput improvement on one of the largest EDA clusters. 75% lower queue wait times via smart queues.
Time to Value
Weeks
The Challenge · 01
Two problems, hiding inside one budget line.
Synaptics' chip-design infrastructure : the systems that run the electronic-design-automation workloads chip teams depend on : was bottlenecked. Job queues backed up. Throughput suffered. The design team lost cycles to waiting rather than shipping designs.
The engineering answer was obvious: smart queues, better workload routing, a modern operational data foundation.
The harder question was budget. A traditional SI proposal would have eaten most of the budget on the discovery and mapping phase : leaving the smart-queue work, the part that actually moved the throughput numbers, in a second phase that might never get funded.
How We Delivered · 02
Aedeon did the discovery. The budget went where the throughput came from.
Aedeon ran source discovery across Synaptics' legacy EDA infrastructure, captured the dependency map, and produced the validation harness in a fraction of the time an analyst tier would have taken. That collapse freed budget : budget that a traditional engagement would have spent on discovery hours. Forward-deployed engineers spent the freed budget on the smart-queue architecture and the cutover work. That is what made the 40% throughput number possible.
Aedeon's Lane
- •Discovery across Synaptics' EDA infrastructure
- •Dependency mapping and lineage extraction
- •Validation harness for parallel-run against legacy
- •Observability instrumentation for the new operational data lake
- •DataOps automation through to production
Forward-Deployed Engineers' Lane
- •Target architecture choice : smart-queue design specific to chip-design workloads
- •Workload prioritization and cutover sequencing
- •Acceptance criteria with Synaptics' design and ops teams
- •Production cutover ownership
- •Cost re-baselining and team handover
In Production · 03
40% more throughput. 75% lower wait times. Design cycles reallocated from waiting to shipping.
Throughput on one of Synaptics' largest EDA clusters improved by up to 40%. Queue contention dropped 75% on the new smart queues.
The design team's cycles reallocated from waiting on jobs to shipping designs : the work Synaptics actually exists to do.
The operational data lake on AWS became the foundation the next initiative builds on.
“We improved throughput on one of our largest EDA clusters by up to 40%, and improved queue contention with 75% lower wait times using smart queues.”
: Michael Brooker, CTO, Synaptics
Why This Matters · 04
Agent-native delivery moves the numbers : even when the product is not an agent.
This is a data-platform story where the customer-facing product is operational data infrastructure, not an AI agent. The throughput number still moved because the delivery model was agent-native.
Aedeon doing the discovery in days rather than weeks is what made the smart-queue engineering work fit inside the budget.
The lesson generalizes: every modernization has a delivery-model cost line. Agent-native delivery moves that cost line, and the savings reallocate into the engineering that actually moves the business outcome.