Scalable AI Acceleration for Intelligent SoC Development
APARAJIT is a licensable AI Accelerator IP developed for AI inference, tensor processing and parallel computing.
Built through indigenous semiconductor research and engineering, APARAJIT enables engineering teams to integrate AI acceleration into custom semiconductor platforms for embedded and intelligent computing applications.
Available in Micro, Mini and Mega configurations, APARAJIT uses a common architecture across the product family, supporting software portability while addressing different deployment requirements. It is designed to operate alongside the AJIT Processor Core IP within custom System-on-Chip (SoC) platforms.
WHY APARAJIT
AI Acceleration Within a Common Architecture
Modern semiconductor platforms increasingly require dedicated acceleration for AI inference, tensor processing and parallel computing.
APARAJIT combines dedicated AI acceleration with programmable processing through its integration with AJIT, providing a common technology foundation across its product configurations.
Scalable Product Family
APARAJIT is available in three configurations:
Micro
1 AJIT Processor Core +
1 APARA AI Accelerator Core
Mini
4 AJIT Processor Cores +
4 APARA AI Accelerator Cores
Mega
64 AJIT Processor Cores +
64 APARA AI Accelerator Cores
Common Architecture
The Micro, Mini and Mega configurations share a common hardware and software architecture, enabling software portability across the product family.
APARA AI ACCELERATOR ARCHITECTURE
Built for AI Inference and Parallel Computing
The APARA AI Accelerator forms the computational foundation of APARAJIT Technology.
These architectural components provide the execution, memory and data-movement capabilities used across the APARAJIT technology platform.
Its documented architecture includes:
MULTIPLE PRECISION SUPPORT
Designed for Different AI Workloads
These formats support AI inference, tensor processing and parallel computing workloads.
The APARA AI Accelerator supports multiple floating-point and integer precision formats:
| Precision | Documented Compute |
|---|---|
| FP4 / INT4 | 256 Ops/Cycle |
| FP8 / INT8 | 128 Ops/Cycle |
| FP16 / INT16 | 64 Ops/Cycle |
| FP32 / INT32 | 16 Ops/Cycle |
| FP64 / INT64 | 8 Ops/Cycle |
APARAJIT Micro
APARAJIT Micro combines 1 AJIT Processor Core with 1 APARA AI Accelerator Core within the common APARAJIT architecture. It is designed for embedded AI, robotics, drones, IoT inferencing and signal processing.
Key Specifications:
- 1 AJIT Processor Core
- 1 APARA AI Accelerator Core
- 5-Constellation GNSS Support
- 4 Independent MIMD Threads
- 8 Hardware Execution Lanes
- 4 Parallel Memory Banks
- Dedicated DMA Engine
- 16 KB Bundled Instruction Cache
- 256-bit Load/Store Interface
APARAJIT Mini
APARAJIT Mini combines 4 AJIT Processor Cores with 4 APARA AI Accelerator Cores within the common APARAJIT architecture. It is designed for autonomous vehicles, tactical communication, smart controllers and embedded industrial applications.
Key Specifications:
- 4 AJIT Processor Cores
- 4 APARA AI Accelerator Cores
- 4 Independent MIMD Threads
- 8 Hardware Execution Lanes
- 4 Parallel Memory Banks
- Dedicated DMA Engine
- 16 KB Bundled Instruction Cache
- 256-bit Load/Store Interface
APARAJIT Mega
APARAJIT Mega combines 64 AJIT Processor Cores with 64 APARA AI Accelerator Cores within the common APARAJIT architecture. It is designed for Edge HPC, real-time telemetry, dense signal processing, portable compute and inference at scale.
Key Specifications:
- 64 AJIT Processor Cores
- 64 APARA AI Accelerator Cores
- 4 Independent MIMD Threads
- 8 Hardware Execution Lanes
- 4 Parallel Memory Banks
- Dedicated DMA Engine
- 16 KB Bundled Instruction Cache
- 256-bit Load/Store Interface
Explore Scalable AI Acceleration for Your Next SoC
Evaluate the APARAJIT AI Accelerator IP family and its Micro, Mini and Mega configurations for your semiconductor development requirements.
APARAJIT + AJIT
Programmable Processing Meets Dedicated AI Acceleration
AJIT Processor Core IP
provides programmable processing and system control.
APARAJIT AI Accelerator
provides dedicated acceleration for AI inference, tensor processing and parallel computing.
Together, they form a unified semiconductor architecture for intelligent SoC development.
AI Acceleration Across Intelligent Computing
Drones
Onboard AI inference and intelligent processing.
Robotics
Programmable processing combined with dedicated AI acceleration.
IoT Inferencing
Local AI inference on embedded IoT devices.
Signal Processing
Acceleration for signal-processing workloads.
The broader Powai Labs portfolio also identifies applications across embedded systems, industrial automation, defence, aerospace, autonomous systems and edge computing.
LICENSABLE SEMICONDUCTOR IP
Designed for Custom SoC Integration
APARAJIT is available as licensable Semiconductor IP for integration into custom System-on-Chip designs.
Its common architecture allows organisations to work across Micro, Mini and Mega configurations while maintaining software portability across the product family.