Uncovering hidden talent through contextual résumé intelligence 

Yonatan Elman, Michael Kovalchuk, Roni Fadlon

GitHub

Traditional résumé parsers depend on explicit keywords, missing many competencies that candidates demonstrate indirectly through their achievements and responsibilities. SkillSight addresses this gap by modeling contextual cues, domain conventions, and action–outcome patterns to infer implicit skills that are not explicitly stated in the résumé. The system transforms unstructured narrative text into a richer skill representation that better reflects a candidate’s true capabilities.

One model. Any system. Accurate triage 

Amit Hazan, Lior Mizrahi, Abed Haj

GitHub

A universal support-ticket triage model that adapts instantly to any software or hardware ecosystem by ingesting the system’s documentation and workflow. It classifies incoming tickets by severity and accurately identifies the underlying components, enabling fast, consistent resolution across diverse systems. 

Clarity in every decision, confidence in every action 

Ohad Biton, Michael Naftalishen, Yossi Okropiridze

GitHub

An AI-driven decision-support system that helps banking clerks determine whether a customer request, transaction, or operational action is permissible under complex and evolving regulatory frameworks. The system interprets multilayered rules, exceptions, and compliance constraints, providing clear, consistent, and auditable guidance for each case. 

Structured evaluation for open-question CS assessments 

Uriel Coehn, Natan Bar, Tomer Bengaev

GitHub

CriteriaGradeLLM is an LLM-based grading system that evaluates computer-science students’ open-question answers using a transparent rubric of 5–10 criteria. It examines conceptual correctness, technical precision, completeness of the response, quality of reasoning, and the relevance of supporting examples. The model generates consistent scores and concise feedback, enabling scalable, fair assessment across large classes while maintaining precise alignment with instructional goals.

Real-time insight into society’s shifting sentiments

Uriel Cohen, Natan Bar, Tomer Bengaev

GitHub

SocioScope is an LLM-driven analysis system that processes large volumes of Twitter content to reveal public sentiment on key social, economic, and political issues such as healthcare, education, employment, and cost of living. By identifying issue-specific aspects and evaluating sentiment toward each, the system provides a structured, real-time view of how populations react to ongoing events and policy discussions. SocioScope helps researchers and decision-makers track opinion trends, detect emerging concerns, and understand the factors shaping public discourse across diverse societal domains. 

Understanding Maritime Urgency Through Text 

Tomer Atia, Ilana Estrin

GitHub

This project builds an automated text-analysis system that classifies maritime messages into distress, security, and routine categories based solely on their content. The model learns linguistic patterns, urgency markers, and domain-specific phrasing typical of maritime communication, enabling it to distinguish life-threatening distress calls from security alerts and routine operational messages. By relying solely on textual cues, the system provides a lightweight, reliable tool for improving message triage and enhancing maritime safety communication workflows. 

Detecting customer emotions to improve support interactions. 

Lior Tsvieli, Yoni Libman, Yuval Reznik

GitHub

EmotiSupport is an emotion-analysis system designed for technical support messages and conversational logs. It interprets the emotional tone behind customer statements such as frustration, confusion, relief, or satisfaction—and links these signals to the progression of the support interaction. By analyzing language patterns common in troubleshooting dialogues, the system highlights emotionally charged moments, identifies escalation risks, and provides actionable insights for improving response strategies and overall service quality. 

From negotiation emails to structured catering deals.

Arkady Doktorovich, Guy Yogev, Hen Mandelbaum

GitHub

CaterExtract analyzes multi-email negotiation threads between clients and catering providers and automatically extracts the key deal parameters. The system converts unstructured, evolving email conversations into a structured deal representation, enabling fast review, comparison, and downstream processing.