Silimate Preqorsor PPA Prediction & SMDB AI Debugger

Silimate is a Mountain View, California startup co-founded by CEO Ann Wu and CTO Akash Levy, featured in Y Combinator's S23 batch, whose team comes from Apple, NVIDIA, Synopsys, AWS and Stanford. It builds AI-native EDA tools and frontend agents that help chip teams reach power, performance and area targets and fix functional bugs in days instead of months.

Preqorsor uses custom machine learning engines that model circuit characteristics to predict and optimise PPA directly from RTL, without waiting for full synthesis runs, and works on designs with millions of gates from block to partition level. SMDB is an AI debugger for root cause analysis of functional bugs. Fine-tuned chip models work with design collateral, and a self-hosted copilot keeps all data on the customer's premises. The company reports more than 10 customers, from startups to Fortune 500 firms.

• Preqorsor: fast machine learning prediction and optimisation of power, performance and area from RTL.
• SMDB: AI debugger performing root cause analysis on functional bugs.
• Fast engines: custom ML models that deterministically capture circuit characteristics.
• Chip models: fine-tuned language models that work with design collateral.
• Scalable analysis: handles designs with millions of gates from block to partition level.
• Self-hosted copilot: on-premise deployment with no data egress.

A Vietnamese RTL design team can use Preqorsor to check power and timing impact of code changes before handing off to synthesis. A verification group can use SMDB to speed up root cause analysis of failing tests. A fabless startup can deploy the self-hosted copilot to keep customer IP inside its own servers.

Customer testimonials on the site report 70x faster PPA optimisation and 11x faster bug resolution. Integration takes less than a day and adapts to process technologies from planar nodes to gate-all-around. On-premise, self-hosted deployment means no design data leaves the customer environment. Early PPA feedback lets designers explore RTL changes before costly synthesis runs.

PPA closure at advanced nodes requires many slow synthesis and implementation iterations, and AI prediction models are emerging to shorten this loop. Chip companies prefer AI tools that run on-premise to protect design IP. Vietnam's expanding front-end design teams, working on projects for global customers, need productivity tools that fit existing secure environments.

Co-founded by Ann Wu (CEO) and Akash Levy (CTO); located on Castro Street, Mountain View, California; featured in Y Combinator S23. In 2025 the company grew ARR more than 20x across 10+ customers, became profitable, was granted a patent and released Preqorsor v1.0 and SMDB. Preqorsor v2.0 with RTL-level dynamic power prediction is planned.