Synapse
Sign in Publish

Browse papers

461 papers

Signaling Competence in Global Remote Technical Hiring: An Open-Source Framework

Tamheen Zehra · Aug 15, 2026

Abstract: The old way of hiring people is not working well when it comes to finding good software engineers who can work from anywhere. This is because the world is changing and more people are working from home.When companies try to hire people from countries it is hard for them to know if the person is really good at their job. This is a problem because companies do not want to hire someone who is not very good. We think that the things people do online like the code they write and the projects they work on can show if they are really good at their job. We also think that how well someone can communicate with others even when they are not in the place is very important. We looked at a lot of data. Found out that the things people do online are a much better way to know if they are good at their job than the school they went to or the companies they used to work for. We also found out that when people work on open-source projects it is a sign that they are very good at their job.We came up with a way for companies to hire people that is based on what they can actually do, not just what they say they can do. This way is better because it is fair and it helps companies find the people for the job.In the end we think that this new way of hiring people will make it easier for companies to find software engineers and will help people get jobs that they are really good, at. It will also help reduce the costs of hiring people and make sure that everyone has a chance of getting a job.

Signaling Competence in Global Remote Technical Hiring: An Open-Source Framework

The Necromancer: An Adversarial Audit Swarm for Vibe-Coded Web Applications

S. Sai Vijay Surya · Aug 14, 2026

Necromancer is a multi-agent tool that adversarially audits Flask and FastAPI apps for crash bugs, risky dependencies, and unfair treatment across demographic groups. Tested against three real third-party repos, it found genuine crash bugs in an unfamiliar FastAPI app — and, in the process, exposed and fixed a false-positive bug in its own dependency checker. The paper argues that this kind of external validation, including catching your own tool's mistakes, is what real testing looks like for software built quickly with AI assistance.

Software SecurityFairness in AIWeb DevelopmentMachine LearningSoftware Testing

Adaptive Multi-Agent UAV Path Planning: From Classical Optimization to Deep Reinforcement Learning

Hamed Raza · Aug 14, 2026

Abstract:Multi-agent UAV path planning in three-dimensional environ- ments with obstacles is a problem of fundamental importance for search-and-rescue, surveillance, and logistics operations. Existing classical optimization methods produce high-quality paths for static environments but fail when obstacles move or missions change mid-flight. We present a systematic progres- sion from a Grey Wolf Optimizer (GWO) baseline to a Deep Reinforcement Learning (DRL) framework, training both DQN and PPO agents in a custom OpenAI Gym environment for 3D UAV navigation. Our key finding: PPO converges 40% faster than DQN under sparse-reward conditions—the realistic setting where agents receive feedback only upon goal completion or collision. We benchmark all three methods on identical test environments, measuring convergence speed, path optimality, and computational cost. The results reveal a fundamental trade- off: classical methods require zero training but cannot adapt; DRL requires upfront training but enables real-time obstacle avoidance. PPO offers the best cost-quality ratio, making it the preferred choice for practical UAV swarm deployments.

Multi-Agent Systems (MAS)Unmanned Aerial Vehicles (UAVs / Drones)Deep Reinforcement Learning (DRL)Motion Planning & Trajectory Optimization

Holo-AI: Interactive 3D Holographic Avatars Driven by Real-time Emotional and Contextual AI

Krishan Yadav · Aug 13, 2026

Krishan Yadav | krishan4.developer@gmail.com | https://ravatar.com/ai-hologram/ Abstract The convergence of volumetric display technology, real-time computer vision, and large-scale generative models presents an opportunity to create interactive three-dimensional holographic avatars capable of perceiving and responding to human emotional states and environmental context. We present Holo-AI, a unified system architecture that integrates multi-modal emotion recognition, contextual reasoning, and real-time volumetric rendering into a single coherent framework. Our approach employs a transformer-based emotion inference engine operating at 30 Hz on fused audio-visual streams, a hierarchical contextual reasoning layer that maintains session-level state, and a differentiable volumetric renderer that produces light-field-consistent holographic output on commodity spatial displays. We demonstrate a proof-of-concept system achieving end-to-end latency of 127 ms (P95), emotional classification accuracy of 91.3% on the DAIC-WOZ benchmark, and real-time rendering at 45 fps on a single NVIDIA RTX 4090. We evaluate the system across telepresence, educational tutoring, and clinical therapeutic support scenarios, reporting subjective user satisfaction scores of 4.2/5.0. We further discuss the computational constraints of current holographic hardware, the ethical implications of real time emotional inference, and a roadmap toward full volumetric display integration.

Holo-AI: Interactive 3D Holographic Avatars

AI-Based Semiconductor Defect Detection Using Edge Computing and Computer Vision

SHARAN N · Aug 10, 2026

Abstract Semiconductor manufacturing demands extremely precise inspection because tiny defects can greatly impact the performance, reliability, and yield of chips. Traditional inspection methods often rely on rule-based image processing, manual checks, or centralized computing, which can struggle to handle the growing volume and complexity of high-resolution images. This paper introduces an AI-powered approach to detect defects in semiconductor wafers by combining computer vision, deep learning, and edge computing. Using a lightweight neural network running directly on edge devices, the system processes and analyzes inspection images locally. It can identify and classify various defect types like particles, scratches, pattern issues, cracks, bridging, and missing structures. By handling the data on-site instead of sending large images to a central server, the system reduces delays, bandwidth needs, and dependency on network reliability. Designed for near-real-time use in manufacturing, this approach aims to deliver fast, accurate defect detection while efficiently using computing resources. We evaluate the system’s performance with metrics such as accuracy, precision, recall, F1-score, and inference speed. Overall, this research shows how combining AI with edge computing offers a scalable and responsive solution to improve semiconductor defect inspection. Keywords: Semiconductor Manufacturing, Defect Detection, Artificial Intelligence, Computer Vision, Edge Computing, Deep Learning, CNN, Wafer Inspection, Machine Vision, Industrial AI.

AI-Based Semiconductor Defect Detection Using Edge Computing and Computer Vision

AI-Based Semiconductor Defect Detection Using Edge Computing and Computer Vision

SHARAN N · Aug 10, 2026

Sharan N, Sapthagiri NPS University, Bengaluru-57 Abstract Semiconductor manufacturing demands extremely precise inspection because tiny defects can greatly impact the performance, reliability, and yield of chips. Traditional inspection methods often rely on rule-based image processing, manual checks, or centralized computing, which can struggle to handle the growing volume and complexity of high-resolution images. This paper introduces an AI-powered approach to detect defects in semiconductor wafers by combining computer vision, deep learning, and edge computing. Using a lightweight neural network running directly on edge devices, the system processes and analyzes inspection images locally. It can identify and classify various defect types like particles, scratches, pattern issues, cracks, bridging, and missing structures. By handling the data on-site instead of sending large images to a central server, the system reduces delays, bandwidth needs, and dependency on network reliability. Designed for near-real-time use in manufacturing, this approach aims to deliver fast, accurate defect detection while efficiently using computing resources. We evaluate the system’s performance with metrics such as accuracy, precision, recall, F1-score, and inference speed. Overall, this research shows how combining AI with edge computing offers a scalable and responsive solution to improve semiconductor defect inspection. Keywords: Semiconductor Manufacturing, Defect Detection, Artificial Intelligence, Computer Vision, Edge Computing, Deep Learning, CNN, Wafer Inspection, Machine Vision, Industrial AI.

AI-Based Semiconductor Defect Detection Using Edge Computing and Computer Vision

Flexible and efficient triboelectric nanogenerators based on PVDF and boron nitride composite yarns and mats.

Sukumaran S et al. · Jul 9, 2026

Flexible and scalable energy-harvesting materials are the driving force behind the emerging era of wearable and self-powered electronics that seamlessly integrate into modern life. Hence, triboelectric nanogenerators (TENGs) offer a versatile solution for integrating energy generation into smart textiles. Here, electrospun poly(vinylidene fluoride) (PVDF) mats and yarns incorporating boron nitride (BN) nanoparticles (1, 3, 5, and 10 wt%) were fabricated and characterized. The 5 wt% BN/PVDF composite exhibited the highest β-phase content and crystallinity, owing to the role of BN as an efficient nucleating agent that facilitates β-phase crystallization through strong interfacial interactions between the nanofiller and PVDF matrix. The triboelectric output was systematically compared across different structural configurations, including electrospun mats, yarns, and rolled-mat geometries. The BN/PVDF yarn-based TENG delivered the highest power density of 303 ± 0.30 mW m -2 , representing ∼113% enhancement over pristine PVDF yarn and superior to previously reported PVDF-based devices. Moreover, despite its smaller active area, the yarn device produced higher power density than the mat counterpart (297 ± 0.43 mW m -2 ). These findings demonstrate that BN incorporation and yarn-based architecture enhanced power generation, providing a scalable route toward high-performance, flexible nanogenerators for wearable and self-powered electronics.

Engineering

Creating molecular complexity in the chemoenzymatic synthesis of chlorothricin analogues using tandem Diels-Alderases.

Devine AJ et al. · Jul 8, 2026

Chlorothricin is a polyketide-derived natural product isolated from Streptomyces antibioticus . It possesses an elaborate pentacyclic aglycone core which incorporates a spirotetronic acid moiety, linked to a trans -decalin system, embedded within a macrocycle. Using synthetic substrate analogues and purified recombinant proteins, here we demonstrate that assembly of this scaffold proceeds via sequential biocatalytic Diels-Alder reactions, promoted by the enzymes ChlE3 and ChlL. Both Diels-Alderases exhibit sufficiently relaxed substrate selectivity to facilitate access to non-natural chlorothricin analogues via biotransformations. The X-ray crystal structure of ChlE3 reveals the molecular basis of decalin formation by this enzyme. Harnessing this enzymatic cascade in biocatalysis could provide a valuable biomimetic route to both natural and non-natural spirotetronates, and the work described herein lays the foundation for application of these enzymes in chemoenzymatic syntheses of complex products.

Biochemistry, Genetics and Molecular Biology

Simultaneous learning of static and dynamic charges.

Stärk P et al. · Jul 8, 2026

Long-range interactions and electric response are essential for accurate modeling of condensed-phase systems, but capturing them efficiently remains a challenge for atomistic machine learning. Traditionally, these two phenomena can be represented by static charges that underlie Coulomb interactions between atoms, and dynamic charges such as atomic polar tensors-aka Born effective charges-describing the response to an external electric field. We critically compare different approaches to learn both types of charges within a single model architecture, taking bulk water and water clusters as paradigmatic examples: (1) learning them independently; (2) coupling static and dynamic charges based on their physical relationship with a single global coupling constant to account for dielectric screening; (3) coupled learning with a local, environment-dependent screening factor. In the coupled case, correcting for dielectric screening is essential, yet the common assumption of homogeneous, isotropic screening breaks down in heterogeneous systems such as water clusters. A learned, environment-dependent screening restores high accuracy for the dynamic charges. However, the accuracy gain over independent dynamic predictions is negligible, while the computational cost increases compared to using separate models for static and dynamic charges. This suggests that, despite the formal connection between the two charge types, modeling them independently is the more practical choice for both condensed-phase and isolated cluster systems.

Materials Science