Intuit Israel

𝐂𝐚𝐧 𝐐𝐮𝐚𝐧𝐭𝐮𝐦 𝐂𝐨𝐦𝐩𝐮𝐭𝐢𝐧𝐠 𝐒𝐭𝐨𝐩 𝐅𝐢𝐧𝐚𝐧𝐜𝐢𝐚𝐥 𝐅𝐫𝐚𝐮𝐝? At the Geektime Code conference earlier this week, we skipped the standard corporate pitch to focus on something far more interesting: how we’re exploring ways in which quantum computing can detect complex Fraud Rings at scale. It’s was great meeting >500 technologists who stopped by out booth, engaged our engineers and asked the tough technical questions. That’s exactly why we love this ecosystem of ours. Racheli Lazar Ofer Rivlin Gleb Keselman Maya Klein Aya Klier Yarden Gur Omer Ziner Ido Elron Or Shemesh

How can hackers manipulate GenAI coding tools to achieve botnet-scale compromise? A new study exposes a paradigm shift in AI vulnerabilities. Until now, most prompt-injection exploits were "push-based", requiring attackers to target victims individually (e.g., via a malicious email), which inherently limited their scale. Enter "HalluSquatting" - the first scalable, "pull-based" supply chain attack. Because AI coding assistants and autonomous agents frequently hallucinate non-existent resource identifiers (like package names, GitHub repo URLs, or skill names), attackers can now: ➡️ Anticipate these hallucinated identifiers using predictive modeling. ➡️ Preemptively register them across public registries. ➡️ Seed them with malicious code or instructions. When LLMs seek out to pulls these hallucinated resources, it triggers automatic remote tool/code execution - giving attackers a hands-free bridge to massive botnet-scale exploits. This research, which has just been featured in Ars Technica, was co-authored researchers from Tel Aviv University and the Technion (Elad Feldman Aya Spira Stav Cohen Ben Nassi, PhD Avishai Wool) alongside our very own Ron Bitton, PhD, Principal AI Security Researcher at Intuit Israel. Read the full breakdown on Ars Technica: https://lnkd.in/eCJSKWh5

Code, collaboration, and carbs. Can't beat that trio. When Global Engineering Days (GED) takes over Intuit Tel Aviv, our engineering, data, and product teams are completely all in. This is where casual coffee chats spark passionate whiteboarding. This is where quick POC demos turn into the actual product roadmap. This is what happens when you empower engineers to build what matters. 🚀

We’ve just been ranked the #2 Best Company to Work For by Great Place to Work – Israel. This not just a badge. It's a recognition that reflects the 𝐬𝐭𝐫𝐨𝐧𝐠 𝐞𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠 𝐜𝐮𝐥𝐭𝐮𝐫𝐞, 𝐡𝐢𝐠𝐡 𝐭𝐫𝐮𝐬𝐭, 𝐚𝐧𝐝 𝐭𝐞𝐜𝐡𝐧𝐢𝐜𝐚𝐥 𝐞𝐱𝐜𝐞𝐥𝐥𝐞𝐧𝐜𝐞 our team built here. And it belongs entirely to the people - both past and present - who shape Intuit Israel. For the full list as published on N12: https://lnkd.in/d255eQkq Jonathan Pietra

Meet Ofir Ben Shoham; Intuit researcher tackling one of the biggest challenges in LLM deployment: inference speed. His newly published paper introduces a principled method for speeding up speculative decoding by trimming the draft model's vocabulary to only what matters for a given domain. Link to paper 🔗 https://lnkd.in/gcQBbdN9 Why it matters: LLMs generate text one token at a time, and every step runs through billions of parameters. That latency drives up costs and limits how many requests a system can handle. The paper addresses the fundamental trade-off: a smaller draft vocabulary reduces draft latency but limits token coverage, thereby reducing the acceptance rate of proposed tokens. This trade-off highlights the need for a principled balance between draft speed and the accuracy of draft proposals. To address this, Ofir’s approach formalizes the coverage–latency trade-off as a constrained optimization problem. The impact: 🔹 16% latency reduction 🔹 20% throughput improvement on domain-specific tasks 🔹 97% smaller draft vocabularies with 93.7% coverage retained For organizations running LLMs in production (think: finance, customer support, code generation), this translates directly to lower infrastructure costs and better user experiences. This research is #opensource — because faster, more efficient LLMs benefit everyone. Great work, Ofir! 👏

What happens when 300 of Israel’s top data minds gather in one room? Absolute magic. ✨ Hosting the 9th annual Women in Data Science (WiDS) conference at Intuit Israel was an incredible reminder of the power of this community. From LLM reward functions to medical NLP and clinical reliability, it was a true deep-dive into the latest developments in AI science. The key takeaways? 🔹Interpretability is the next frontier for LLM reliability 🔹The role of the Data Scientist is evolving from model-builder to strategic architect. 🔹Building for production requires more than just a good prompt — it's about robust engineering. Huge thanks to our amazing ambassadors Yael Hochma, PhD, Nitzan Gado and the 300+ researchers and engineers who joined us to push the boundaries of what’s possible in AI. See you next year! 🚀 #WiDSTLV #IntuitIsrael #DataScience #GenAI #WomenInTech #MachineLearning

שימו לב- נותרו מקומות אחרונים!♟️🏆 לאור המצב החלטנו להאריך את ההרשמה לאליפות ההייטק הקבוצתית בשחמט בעוד מספר שבועות, כשכמעט 80 חברות כבר אישרו השתתפות! לפרטים נוספים והרשמה צרו קשר כאן 👇 אלון: 054-3517190 מייל: Alonmindlin1@gmail.com TechMate-The Ultimate Move | Intuit Israel | Zscaler

I’m biased, but this is a genuinely great place to work 🤩 We’re looking for a design leader to join our IDX (Intuit Data Exchange) team in Intuit Israel. It’s a high-impact role focused on complex systems and AI-driven experiences. If you have 12+ years of experience and enjoy a "startup culture, big tech scale" environment, I’d love to chat. We’re hybrid, value craft, and have a solid team of designers here. Interested or know someone? Send me a message. Full job description here: https://lnkd.in/dJWcrsJz