Arun Ravindran

Large language models (LLMs) are emerging as powerful tools for hardware design, with recent work exploring their ability to generate register-transfer level (RTL) code directly from natural-language specifications. This paper presents a critical review and empirical evaluation of LLM-based RTL generation. We examine thirty-one published efforts, classifying their use of techniques such as fine-tuning, reinforcement learning, retrieval-augmented prompting, and multi-agent orchestration across eight methodological dimensions including debugging support, post-RTL metrics, and benchmark development. Building on this synthesis, we experimentally evaluate frontier commercial and open-weight models—GPT-4.1, GPT-4.1-mini, Claude Sonnet 4, and Llama 4 Maverick —on the VerilogEvalV2 and RTLLMv2.0 benchmarks under both single-shot generation and a lightweight ReAct-style reflection loop, with compilation and simulation performed through Icarus Verilog interfaced via the Model Context Protocol (MCP). Results show that these models achieve up to 89.74% on VerilogEval and 96.08% on RTLLM, matching or exceeding prior domain-specific pipelines without specialized fine-tuning. Detailed failure analysis reveals systematic error modes, including FSM mis-sequencing, handshake drift, blocking vs. non-blocking misuse, and state-space oversimplification. Finally, we outline a forward-looking research roadmap toward natural-language-to-System-on-Chip (SoC) design, emphasizing realistic benchmarks and open flows, richer specification formalisms, AI–human collaborative design environments, and system-level feedback that spans physical design, firmware, and design space exploration. Together, this work provides both a synthesis of recent advances and a baseline evaluation of frontier LLMs, highlighting opportunities and challenges in moving toward AI-native electronic design automation.

Large language models (LLMs) are emerging as powerful tools for hardware design, with recent work exploring their ability to generate register-transfer level (RTL) code directly from natural-language specifications. This paper provides a survey and evaluation of LLM-based RTL generation. We review twenty-six published efforts, covering techniques such as fine-tuning, reinforcement learning, retrieval-augmented prompting, and multi-agent orchestration, and we analyze their contributions across eight methodological dimensions including debugging support, post-RTL metrics, and benchmark development. Building on this review, we experimentally evaluate frontier commercial models---GPT-4.1, GPT-4.1-mini, and Claude Sonnet 4---on the VerilogEval and RTLLM benchmarks under both single-shot and lightweight agentic settings. Results show that these models achieve up to 89.74% on VerilogEval and 96.08% on RTLLM, matching or exceeding prior domain-specific pipelines without specialized fine-tuning. Detailed failure analysis reveals systematic error modes, including FSM mis-sequencing, handshake drift, blocking vs. non-blocking misuse, and state-space oversimplification. Finally, we outline a forward-looking research roadmap toward natural-language-to-SoC design, emphasizing controlled specification schemas, open benchmarks and flows, PPA-in-the-loop feedback, and modular assurance frameworks. Together, this work provides both a critical synthesis of recent advances and a baseline evaluation of frontier LLMs, highlighting opportunities and challenges in moving toward AI-native electronic design automation.

Modern SoC datapaths include deeply pipelined, domain-specific accelerators, but their RTL implementation and verification are still mostly done by hand. While large language models (LLMs) exhibit advanced code-generation abilities for programming languages like Python, their application to Verilog-like RTL remains in its nascent stage. This is reflected in the simple arithmetic and control circuits currently used to evaluate generative capabilities in existing benchmarks. In this paper, we introduce ArchXBench, a six-level benchmark suite that encompasses complex arithmetic circuits and other advanced digital subsystems drawn from domains such as cryptography, image processing, machine learning, and signal processing. Architecturally, some of these designs are purely combinational, others are multi-cycle or pipelined, and many require hierarchical composition of modules. For each benchmark, we provide a problem description, design specification, and testbench, enabling rapid research in the area of LLM-driven agentic approaches for complex digital systems design. Using zero-shot prompting with Claude Sonnet 4, GPT 4.1, o4-mini-high, and DeepSeek R1 under a pass@5 criterion, we observed that o4-mini-high successfully solves the largest number of benchmarks, 16 out of 30, spanning Levels 1, 2, and 3. From Level 4 onward, however, all models consistently fail, highlighting a clear gap in the capabilities of current state-of-the-art LLMs and prompting/agentic approaches.

The increasing reliance on web services has led to a rise in cybersecurity threats, particularly Cross-Site Scripting (XSS) attacks, which target client-side layers of web applications by injecting malicious scripts. Traditional Web Application Firewalls (WAFs) struggle to detect highly obfuscated and complex attacks, as their rules require manual updates. This paper presents a novel generative AI framework that leverages Large Language Models (LLMs) to enhance XSS mitigation. The framework achieves two primary objectives: (1) generating sophisticated and syntactically validated XSS payloads using in-context learning, and (2) automating defense mechanisms by testing these attacks against a vulnerable application secured by a WAF, classifying bypassing attacks, and generating effective WAF security rules. Experimental results using GPT-4o demonstrate the framework's effectiveness generating 264 XSS payloads, 83% of which were validated, with 80% bypassing ModSecurity WAF equipped with an industry standard security rule set developed by the Open Web Application Security Project (OWASP) to protect against web vulnerabilities. Through rule generation, 86% of previously successful attacks were blocked using only 15 new rules. In comparison, Google Gemini Pro achieved a lower bypass rate of 63%, highlighting performance differences across LLMs.

The increasing reliance on web services has led to a rise in cybersecurity threats, particularly Cross-Site Scripting (XSS) attacks, which target client-side layers of web applications by injecting malicious scripts. Traditional Web Application Firewalls (WAFs) struggle to detect highly obfuscated and complex attacks, as their rules require manual updates. This paper presents a novel generative AI framework that leverages Large Language Models (LLMs) to enhance XSS mitigation. The framework achieves two primary objectives: (1) generating sophisticated and syntactically validated XSS payloads using in-context learning, and (2) automating defense mechanisms by testing these attacks against a vulnerable application secured by a WAF, classifying bypassing attacks, and generating effective WAF security rules. Experimental results using GPT-4o demonstrate the framework's effectiveness generating 264 XSS payloads, 83% of which were validated, with 80% bypassing ModSecurity WAF equipped with an industry standard security rule set developed by the Open Web Application Security Project (OWASP) to protect against web vulnerabilities. Through rule generation, 86% of previously successful attacks were blocked using only 15 new rules. In comparison, Google Gemini Pro achieved a lower bypass rate of 63%, highlighting performance differences across LLMs.

The widespread adoption of web services has heightened exposure to cybersecurity threats, particularly SQL injection (SQLi) attacks that target the database layers of web applications. Traditional Web Application Firewalls (WAFs) often fail to keep pace with evolving attack techniques, necessitating adaptive defense mechanisms. This paper introduces a novel generative AI framework designed to enhance SQLi mitigation by leveraging Large Language Models (LLMs). The framework achieves two primary objectives: (1) generating diverse and validated SQLi payloads using in-context learning, thereby minimizing hallucinations, and (2) automating defense mechanisms by testing these payloads against a vulnerable web application secured by a WAF, classifying bypassing attacks, and constructing effective WAF security rules through generative AI techniques. Experimental results using the GPT-4o LLM demonstrate the framework’s efficacy: 514 new SQLi payloads were generated, 92.5% of which were validated against a MySQL database and 89% of which successfully bypassed the ModSecurity WAF equipped with the latest OWASP Core Rule Set. By applying our automated rule-generation methodology, 99% of previously successful attacks were effectively blocked with only 23 new security rules. In contrast, Google Gemini-Pro achieved a lower bypass rate of 56.6%, underscoring performance variability across LLMs. Future work could extend the proposed framework to autonomously defend against other web attacks, including Cross-Site Scripting (XSS), session hijacking, and specific Distributed Denial-of-Service (DDoS) attacks.

The paper addresses advancements in Generative Artificial Intelligence (GenAI) and digital chip design, highlighting the integration of Large Language Models (LLMs) in automating hardware description and design. LLMs, known for generating human-like content, are now being explored for creating hardware description languages (HDLs) like Verilog from natural language inputs. This approach aims to enhance productivity and reduce costs in VLSI system design. The study introduces "AiEDA", a proposed agentic design flow framework for digital ASIC systems, leveraging autonomous AI agents to manage complex design tasks. AiEDA is designed to streamline the transition from conceptual design to GDSII layout using an open-source toolchain. The framework is demonstrated through the design of an ultra-low-power digital ASIC for KeyWord Spotting (KWS). The use of agentic AI workflows promises to improve design efficiency by automating the integration of multiple design tools, thereby accelerating the development process and addressing the complexities of hardware design.

The widespread adoption of web services has heightened exposure to cybersecurity threats, particularly SQL Injection (SQLi) attacks that target the database layers of web applications. Traditional Web Application Firewalls (WAFs) often fail to keep pace with evolving attack techniques, necessitating adaptive defense mechanisms. This paper introduces a novel generative AI framework designed to enhance SQLi mitigation by leveraging large language models (LLMs). The framework achieves two primary objectives: (1) generating diverse and validated SQLi payloads using in-context learning, thereby minimizing hallucinations, and (2) automating defense mechanisms by testing these payloads against a vulnerable web application secured by a WAF, classifying bypassing attacks, and constructing effective WAF security rules through generative AI techniques. Experimental results using the GPT-4o LLM demonstrate the framework’s efficacy: 514 new SQLi payloads were generated, 92.5% of which were validated against a MySQL database, and 89% of which successfully bypassed the ModSecurity WAF equipped with the latest OWASP Core Rule Set. By applying our automated rule-generation methodology, 99% of previously successful attacks were effectively blocked with only 23 new security rules. Results with the Google Gemini-Pro LLM are also presented.

The falling cost of IoT cameras, the advancement of AI-based computer vision algorithms, and powerful hardware accelerators for deep learning have enabled the widespread deployment of surveillance cameras with the ability to automatically analyze streaming video feeds to detect events of interest. While streaming video analytics is currently largely performed in the cloud, edge computing has emerged as a pivotal component due to its advantages of low latency, reduced bandwidth, and enhanced privacy. However, a distinct gap persists between state-of-the-art computer vision algorithms and the successful practical implementation of edge-based streaming video analytics systems. This paper presents a comprehensive review of more than 30 research papers published over the last 6 years on IoT edge streaming video analytics (IE-SVA) systems. The papers are analyzed across 17 distinct dimensions. Unlike prior reviews, we examine each system holistically, identifying their strengths and weaknesses in diverse implementations. Our findings suggest that certain critical topics necessary for the practical realization of IE-SVA systems are not sufficiently addressed in current research. Based on these observations, we propose research trajectories across short-, medium-, and long-term horizons. Additionally, we explore trending topics in other computing areas that can significantly impact the evolution of IE-SVA systems.

The falling cost of cameras, the advancement of AI based computer vision algorithms, and powerful hardware accelerators for deep learning have enabled wide-spread deployment of surveillance cameras with the ability to automatically analyze streaming video feeds to detect events of interest. While streaming video analytics is currently largely done in the cloud, edge computing has emerged as a pivotal component due to its advantages of low latency, reduced bandwidth, and enhanced privacy. However, a distinct gap persists between the state-of-the-art computer vision algorithms, and successful practical implementation of edge-based streaming video analytics systems. This paper presents a comprehensive review of more than 30 research papers published over the last 6 years on edge video analytics systems. The papers are analyzed across 17 distinct dimensions. Unlike prior reviews, we examine each system holistically, identifying their strengths and weaknesses in diverse implementations. Our findings suggest that certain critical topics necessary for the practical realization of edge video analytics systems are not sufficiently addressed in current research. Based on these observations, we propose research trajectories across short, medium, and long term horizons. Additionally, we explore trending topics in other computing areas that can significantly impact the field of edge video analytics.