Elaheh Rashedi

TIL that Iran launched Ballistic missiles at The Weizmann Institute of Science in June 2025, destroying labs and years of work. I don't know how you justify targeting Research Campuses. I was looking for the best places to conduct SOTA research in computer science. I'm waiting on my Master's until these wars subside, but this is so not cool. No student or researcher should have to factor "will my university get bombed?" into their career planning. That should be a basic global consensus. For those that don't know, The Weizmann Institute of Science offers top notch MSc. Programs free of cost so that students can completely devote their time to Scientific Research and Study. I was getting excited right before this news infuriated me. To the students and researchers at Weizmann: I’m deeply sorry for your loss. Years of work should not be erased by a missile salvo. The least the rest of us can do is not look away, and keep insisting that universities must be off‑limits everywhere.

1

🚀 Excited to share that our latest research paper, ‘AI-driven fuzzy requirement engineering for cybersecurity: integrating linguistic decision models,’ has been published in the Iran Journal of Computer Science (SCIE). In this work, we propose a novel methodology that combines Artificial Intelligence (AI) and fuzzy logic to tackle the complexities of cybersecurity requirement engineering. A big thank you to my co-authors, Dr. Shahnawaz Ahmad, Dr. Mohd. Arif, and Dr. Mohd. Aquib Ansari for this collaboration. 👉 Read the full paper here: https://lnkd.in/gg8bxdAK #AI #Cybersecurity #FuzzyLogic #RequirementsEngineering #ChatGPT #ResearchPublication

66

8 Comments

Building a Production-Ready RAG System for Persian Language Processing I’m excited to share insights from developing an advanced Retrieval-Augmented Generation system specifically engineered for Persian/Farsi document processing, leveraging cutting-edge ML infrastructure and architectural patterns. 🔬 Technical Architecture Highlights: Advanced RAG Implementation: ∙ Self-Corrective Agentic RAG with adaptive query rewriting and iterative refinement ∙ Hybrid search architecture combining dense embeddings with SPLADE sparse representations ∙ Multi-agent orchestration enabling context-aware document interaction and general conversation modes Production Infrastructure: ∙ NVIDIA Triton Inference Server deployment on H100 GPUs with vLLM backend ∙ Milvus vector database for high-performance similarity search ∙ PostgreSQL for structured data with MinIO object storage for scalable file management ∙ Model ensemble: Qwen2.5-72B-Instruct-AWQ, DeepSeek-OCR, and sentence-transformers ML Pipeline Innovations: ∙ Automated synthetic QA pair generation using LangGraph workflows ∙ Cross-encoder reranking (ms-marco-MiniLM-L6-v2) for precision optimization ∙ Dynamic prompt management system with A/B testing capabilities and versioning ∙ Fine-tuning pipeline with quality-scored synthetic datasets Engineering Principles: ∙ Clean Architecture with dependency injection and separation of concerns ∙ Docker containerization for reproducible deployments ∙ Comprehensive observability with metrics tracking and performance monitoring This project demonstrates how to architect enterprise-grade RAG systems that combine theoretical ML advancements with practical production considerations, particularly for low-resource languages. Open to discussing technical implementation details, architectural decisions, or challenges in deploying LLMs at scale. #MLEngineering #RAG #NLP #LLMs #VectorSearch #ProductionML #AI #DeepLearning #Persian #TritonInference #CleanArchitecture

1

Building a Fully Local Speech-to-Text & Refinement Pipeline for Persian I recently implemented a production-ready, on-prem AI pipeline for Persian speech processing. The goal wasn't just a "cool demo"—it was to achieve high-fidelity, practical correctness for administrative and legal workflows. The Challenge: Raw ASR output (even from large models) is often phonetically accurate but structurally weak. By decoupling Recognition from Linguistic Refinement, we achieve results that are actually usable in professional documentation. Architecture Overview: 1) ASR Layer: Whisper (Large/Medium) via FFmpeg normalization. We use deterministic decoding ($temp=0$) to ensure output stability. 2) Normalization: A rule-based hybrid approach to handle Arabic-to-Persian character conversion and artifact removal. 3) Refinement Layer: A local LLM (Qwen-Instruct 7B–14B) served via Ollama. It transforms raw transcripts into formal, administrative Persian while strictly preventing hallucinations. 4) Service Layer: A robust Flask REST API handling multipart uploads and persistent JSON artifacts. Design Principles: 🔒 Data Sovereignty: 100% local inference for sensitive domains. 🎯 Precision-First: Constrained ASR + LLM for deterministic outcomes. 🧱 Modular Design: Clear separation between ASR and language editing. Practical Applications: This system is designed specifically for complaint/report dictation and legal drafting where privacy and syntactic structure are non-negotiable. Happy to discuss trade-offs—vLLM vs. Ollama, quantization strategies, or prompt engineering for Persian—with others building on-prem NLP systems. #AIEngineering #SpeechRecognition #PersianNLP #OnPremAI #Whisper #Qwen #LLMSystems #MLOps

6

Bigger datasets aren’t always better MIT LIDS researchers have developed a new way to pinpoint exactly how much data is needed to solve complex problems. LIDS PhD student Omar Bennouna, former MIT postdoc Mohammed Amine Bennouna, and LIDS PIs Saurabh Amin and Asu Ozdaglar have introduced an algorithmic method that provably identifies the smallest dataset required to guarantee an optimal solution—often using far fewer measurements than conventional approaches assume. Their mathematical framework applies broadly to structured decision-making under uncertainty, from supply chain management to electricity network optimization. “We’ve shown that with careful selection, you can guarantee optimal solutions with a small dataset—and we provide a method to identify exactly which data you need,” says Asu Ozdaglar. Learn more and read the paper: https://bit.ly/49sqkNI MIT EECS MIT Civil and Environmental Engineering MIT Institute for Data, Systems, and Society (IDSS) MIT Schwarzman College of Computing MIT School of Engineering

100

2 Comments

🎓 Session 5 of Machine Learning for Teens (in Persian) is now live! 📹 Watch the video here https://lnkd.in/dSz9h2dd In this session, I walk through key ML concepts using two practical datasets: housing.csv and the digits dataset from scikit-learn. While the breast cancer dataset is part of the assignment, this video focuses on hands-on implementation with real data. Covered topics include: ✅ What causes overfitting (high variance) ✅ Variance formula and visualization ✅ Regularization: L1, L2, ElasticNet ✅ Cross-validation: K-Fold vs StratifiedKFold ✅ Balanced vs imbalanced datasets ✅ Why cross-validation doesn’t prevent overfitting directly ✅ .score() vs accuracy_score() — what’s the difference? This is part of my Machine Learning for Teens series — designed to make ML concepts reproducible, visual, and beginner-friendly for Persian-speaking learners. 📌 I’ll post an English summary in the comments for non-Persian speakers. #MachineLearning #MLforTeens #DataScience #Overfitting #Regularization #CrossValidation #Python #ScikitLearn #LogisticRegression #ImbalancedData #ElasticNet #MLEducation #PersianEducation

4

4 Comments

#Tsinghua_University researchers have developed a new #shortest-path algorithm that is faster than #Dijkstra’s method, which has been the standard for 40 years! Unlike Dijkstra’s algorithm, which relies on sorting nodes, their approach groups certain nodes and explores more selectively, reducing the need for sorting. This breakthrough works even on graphs with any weights and recently won the Best Paper award at STOC. #new-shortest-path-algorithm

16

2 Comments

AI in our lifetimes and at this moment is on the brink of becoming the most consequential thing in human history in comparison to all transformative predecessors (axes, fire, wheels, language, writing, guns, steam engines, factories, oil, electricity, cars, trains, planes, and computers) combined. In this TED Talk, Mustafa Suleyman, CEO of AI at Microsoft, offers an insightful answer to the question "What is an AI anyway?" Working his way to his take on an answer, he considers the consistently inevitable Pandora's Box of positive and negative consequences of all other global societal sweeping transformative discoveries and exhorts his audience on the importance of taking stock and caution for the consequential mix that will accompany AI's maturation. https://lnkd.in/eNQ2BGgX

1

✨ Imagine this: You’re talking to an AI, and instead of forgetting everything after the chat ends, it remembers. It recalls past conversations, learns from them, and builds context like a trusted partner. That’s precisely the world I stepped into with the LLMs as Operating Systems: Agent Memory course from DeepLearning.AI. . 🚀 At first, it felt almost sci-fi: The LLM’s context window became its short-term memory. External storage turned into long-term memory. And the agent itself? It was like an operating system, intelligently deciding what to keep, what to forget, and when to retrieve information. 🧠 The highlight for me was experimenting with MemGPT-style agents. These aren’t just chatbots — they can self-edit their memory, update knowledge, and even collaborate with other agents using shared memory. Suddenly, “AI memory” wasn’t just storage — it was strategy. And the “aha!” moment? When I saw how selective memory works. It’s not just about remembering — it’s about remembering responsibly like an AI that recalls patterns but knows when to redact personal details depending on security or authority. That shift made me realize that memory is the key to trusting AI. This course didn’t just teach me a framework — it gave me a glimpse into the future of AI agents as true operating systems. Agents that don’t just answer, but grow with you. I walked away thinking: 👉 The difference between a chatbot and a real agent isn’t just intelligence — it’s memory. And now, I can’t stop wondering — how will memory-driven agents reshape industries like healthcare, education, or even project management? What do you think — are we ready for AI that truly remembers? #AI #AgenticAI #LLMs #ArtificialIntelligence #MemGPT #FutureOfWork

25

1 Comment

Today’s guide by Reza Bahmanzadeh, AI Engineer on our Professional Services Team, teaches the essentials of serving large models with vLLM: https://lnkd.in/eHK9nwZm It builds from foundational neural network concepts, like transformers and attention, to introduce practical inference workflows, explore useful vLLM features and offer hands-on, operational guidance for efficient deployments. #LLMs #Serving #Inference #TechGuides

35

1 Comment