At a glance
| Booth | P6117 |
| Country | SG |
| Website | www.sixsense.ai |
Company profile
SIXSENSE PTE. LTD. is a purpose-built AI platform for fabs and OSATs, covering automated defect inspection, lot disposition and defect root-cause analysis. The company's mission is to give every chip factory a brain, reflecting a view that semiconductor factories should see details, learn from production signals and make better decisions in real time. Its description of semiconductor manufacturing emphasizes thousands of process steps, hundreds of machines and a production environment with no margin for error. SIXSENSE PTE. LTD. began in 2018 with a vision of bringing trustworthy, intelligent automation to semiconductor manufacturing. The AI platform transforms raw production data into intelligence intended to catch hidden issues early, improve yield and provide a greater level of manufacturing control. The company describes a workplace model in which employees can take ownership of impactful projects and work on difficult technical challenges from the start. Its employee benefits description includes competitive salaries, benefits and 100% health insurance for employees and their families. Global-exposure opportunities include travel to Singapore, Taiwan, Malaysia, the US and Israel for customer pilots, product discovery and collaboration. Learning and development support includes courses, certifications, fitness programs and personal-development initiatives. The company also describes team off-sites, board-game activities and weekend getaways as part of its community practices. A hybrid working model combines office culture in Singapore and Bengaluru with remote-work flexibility. The global footprint lists locations in Singapore, San Francisco in the USA, Bengaluru in India and Hsinchu County in Taiwan. SIXSENSE PTE. LTD. also states that its products are trusted by top semiconductor companies and other industry leaders across Singapore, Taiwan, Malaysia, China, Israel and the US.
Exhibits
The AI suite is presented as a way to unlock 95% of a fab's hidden intelligence while predicting and preventing costly excursions, cutting scrap costs, reducing cycle time by 30% and scaling output 1.3x without new CAPEX. AI-ADC is the company's end-to-end solution for automating visual inspection and includes tools for model training, deployment and maintenance. Its real-time vision engine analyzes every captured image and classifies suspected defects into hundreds of types with over 99% accuracy and zero escapee. AI-ADC converts defect images into meaningful insights at line speed, reducing review queues, freeing clean-room headcount and providing a foundation for downstream analytics. The platform is described as enabling fabs to scale without depending on additional hiring, training or human-driven quality slips. AI classifies every suspected defect 24/7, providing round-the-clock inspection coverage while reducing the required footprint in clean rooms. Instant AI pass/fail decisions are intended to keep lots moving and save hours at each review step, adding up to days of savings for the fab. By catching tiny critical defects early, the AI allows engineers to correct the process before scrap accumulates and also avoids unnecessary over-rejection from cautious manual screening. Consistent bias-free checks are presented as a way to stop bad wafers from moving downstream and reduce reworks, scraps, RMAs and customer complaints from field failures. The company states that AI can perform work previously handled by dozens of operators, allowing people to focus on higher-value engineering and improvement tasks. Build AI provides an end-to-end environment for preparing training data and building high-accuracy AI models with automatic sampling of useful images and minimal manual effort. Build AI can select a diverse, balanced and de-duplicated data set that includes hard-to-find examples from millions of images, and it supports AI-assisted label correction at 10x speed. Its training approach is described as launching with 0% escapee in production while reducing repeated retraining and rollbacks. Build AI uses true classification models rather than patched detection or anomaly workarounds, avoids bounding-box preparation and targets high accuracy even on extremely small and similar-looking defects. The platform includes large vision foundation models specialized for semiconductors and pre-trained on millions of wafer-defect images. Its stated performance includes more than 95% automation, 0% UR and up to less than 0.05% OR while using image counts in two digits rather than four digits per class. Explainable AI features let users inspect what the model recognizes as a defect and verify what the AI has learned before launching it in production. ClassifAI runs AI models in production across AOI tools and is designed for high-throughput, low-latency visual inspection with near-zero escapes. A single ClassifAI model can cover hundreds of devices and tools across generations, with built-in support for multi-view, multi-tool and multiple magnifications. ClassifAI includes ready integrations for Rescans, Klarf variants and incoming defects and is described as deployable by yield engineers. Its operator interface can auto-grade and funnel new defects to a built-in manual-review interface for 1-click retraining, while interactive wafer maps, hotkeys and color-coded defect maps support operator adoption. VerifAI automatically monitors and adapts production models with drift detection and self-improvement capabilities intended to maintain reliable production performance. Lot Disposition and Root-Cause Analysis highlights high-risk wafers according to defect severity and yield impact so teams can prioritize reviews, recover yield and reduce escapes and scrap. Built on AI-ADC labels, Lot Disposition automatically decides pass, hold, rework or scrap for each lot based on yield impact, recurrence patterns and defect severity. The expanded Lot Disposition workflow writes the verdict into production systems to remove manual sign-offs, avoid costly excursions and keep cycle time lean at peak throughput. Dispose AI surfaces critical repeating defects, wafer-map signatures and yield outliers, supports auto-lot checkout decisions based on yield criteria and low kill-ratio defects, and can place lots with low first-pass or final yield on hold for review. Always-on AI agents correlate classified defects with tool, recipe and chamber data across hundreds of steps, flag parameters drifting out of spec, quantify yield exposure and surface the quickest fix. The company states that analysis that once took weeks can be reduced to minutes through this combination of classified defects and process-context data. CausifAI pinpoints key failure sources by correlating process-tool, lot-history and inspection data, quantifies wafer exposure through wafer-stack analytics and provides reports and charts for yield engineers to trace sources of true and false yield loss. Predict AI forecasts when, where and what to inspect by learning defect likelihood across time, layers and lots. Those predictions can trigger proactive adjustments to inspection sampling and tools before faults occur, with the stated aims of reducing scrap and downtime and moving fabs toward self-optimizing production. Predict AI can forecast die-failure probabilities from inspection, metrology and process data in early steps. It can identify layers, tools and patterns driving repeat defects and dynamically focus sampling on higher-risk areas. Predict AI can also trigger proactive tool and recipe adjustments before scrap starts and reduce unplanned downtime through self-optimizing, data-driven production. Across fabs and OSAT lines, the AI suite automates defect review at scale, traces recurring faults to root causes and immediately flags at-risk wafers, with stated outcomes of lower yield exposure, tighter quality control, higher productivity and higher confidence on each run. AI-ADC is also described as delivering faster accurate defect classification with fewer labeled samples, reducing both cycle time and cleanroom footprint.
Capabilities and products
- semiconductor AI automated defect inspection
- AI-ADC visual-inspection automation
- defect-classification AI model building
- production defect classification across AOI tools
- AI model drift monitoring and self-improvement
- automated lot pass/hold/rework/scrap disposition
- semiconductor defect root-cause analysis
- AI predictive inspection and yield-risk prediction