
Closed
Posted
Paid on delivery
We need an experienced AI engineer or small team to build a production conversational assistant for a university website in Saudi Arabia. The assistant answers questions from students, applicants, parents, staff and visitors in Arabic and English, by text and voice, 24/7. Scale: about 200,000 queries/month, 20,000 users, and 2,000+ concurrent sessions at peak. What you will build RAG question answering: hybrid retrieval, re-ranking, multilingual embeddings, and metadata filters. Every answer cites an official source page. Zero fabrication: claim-level grounding, a faithfulness check, abstention when unsure, and deterministic answers for fees, deadlines and admission criteria from structured data. Automated content ingestion: incremental crawler (JS rendering, change detection), layout-aware PDF and table parsing with OCR, a versioned index with rollback, and updates live within 24 hours. Agentic layer: multi-agent orchestration on an open-source, self-hosted framework (e.g. LangGraph). MCP-style typed tools for ticketing, contact lookup and visit scheduling, with guardrails and fallback to plain retrieval. Voice: streaming speech-to-text and text-to-speech with barge-in, under 2 seconds end to end. Admin portal: dashboards, conversation review and correction workflow, role-based access, audit log, and content approval. Safety: protection against prompt injection and jailbreaks (OWASP LLM Top 10), PII redaction, and scope control. Evaluation: an automated test harness (faithfulness, precision, recall) and red-team results. Delivery: a responsive, RTL-correct widget (WCAG 2.1 AA), Kubernetes deployment, CI/CD, documentation, and full source code handover. Hard constraints All data and models must stay inside Saudi Arabia. There are no external API calls at runtime, so no OpenAI or similar. Open-weight Arabic-capable models are self-hosted. Alignment with Saudi PDPL and NCA controls. Required experience: production RAG with Arabic, LLM serving, Kubernetes, voice pipelines, and agent frameworks. Please send examples of Arabic NLP or RAG work you have shipped. To apply, please include: a proposed architecture in a few lines, your timeline, your rate by milestone, and two relevant references. An NDA is required before detailed documents are shared.
Project ID: 40799345
212 proposals
Remote project
Active 19 hours ago
Set your budget and timeframe
Get paid for your work
Outline your proposal
It's free to sign up and bid on jobs
212 freelancers are bidding on average $6,933 USD for this job

—>> AI-Powered University Chatbot – Arabic & English <<— Hello, I checked JD and you need a secure, self-hosted bilingual AI assistant for a Saudi university, supporting text and voice with accurate, source-cited answers and high concurrency. Key Features: * Arabic/English multilingual RAG with hybrid search and reranking. * Self-hosted open-weight LLMs with no external runtime API calls. * Automated website, PDF and OCR content ingestion. * LangGraph multi-agent orchestration and secure MCP-style tools. * Streaming voice interaction with interruption support. * Admin dashboard, analytics and content approval. * PDPL/NCA-aligned security, PII protection and audit logs. * Automated evaluation, Kubernetes deployment and CI/CD Let’s chat… Thanks.
$5,100 USD in 20 days
9.7
9.7

Hi, This is Elias from Miami. I checked your project description and understand you’re looking to build a conversational AI assistant for a university setting. This involves creating an engaging, user-friendly assistant that can effectively communicate with students and faculty. I’ve developed several AI-powered applications and have a solid grasp of the technical challenges involved in building conversational agents. I’d love to discuss the details and suggest the best approach to ensure your vision is realized effectively. I have a few questions to get a better understanding: Q1 – What specific functionalities do you envision for the AI assistant? Q2 – Are there any existing systems or platforms that the assistant needs to integrate with? Q3 – What user roles do you anticipate, and how do you see them interacting with the assistant? Looking forward to hearing from you.
$8,000 USD in 7 days
8.9
8.9

Hi there, We can help build a secure, bilingual AI university assistant with Arabic and English support, designed for high traffic and reliable, source-grounded answers. Our team has experience with Python, AI/ML, RAG systems, LLM integration, backend development, APIs, and cloud infrastructure. We can work toward a self-hosted architecture using open-weight models, hybrid search, vector databases, LangGraph, and Kubernetes, with a strong focus on data privacy, security, and Saudi PDPL requirements. We would be happy to sign an NDA before reviewing your detailed requirements and discussing the architecture, milestones, and timeline. Could you confirm whether you already have Saudi-based infrastructure available for hosting the models and services? Please check our profile reviews and completed projects. Send me a message to discuss this project further. Best regards, Asif Ali Zaman
$5,000 USD in 22 days
8.5
8.5

Hello, We carefully studied the description of your project and we can confirm that we understand your needs and are also interested in your project. Our team has the necessary resources to start your project as soon as possible and complete it in a very short time. We are 25 years in this business and our technical specialists have strong experience in PHP, Java, JavaScript, AngularJS, Kubernetes, Natural Language Processing, AI Chatbot Development, AI Development, AI Voice Agents, AI Automation and other technologies relevant to your project. Please, review our profile https://www.freelancer.com/u/tangramua where you can find detailed information about our company, our portfolio, and the client's recent reviews. Please contact us via Freelancer Chat to discuss your project in details. Best regards, Sales department Tangram Canada Inc.
$7,354 USD in 5 days
8.9
8.9

⭐⭐⭐⭐⭐ Build a Bilingual Conversational Assistant for University Site ❇️ Hi My Friend, I hope you are doing well. I checked all of your project requirements and I can see you are looking for a production AI assistant. You have no need to look any further as Zohaib is here to help you! My team is already doing 50+ similar projects for multilingual RAG assistants. I will build a secure self‑hosted RAG system with Arabic‑English embeddings, voice I/O and an admin portal, all inside Saudi data centers. I will also add automated crawling, PDF parsing and a test harness at no extra cost. ➡️ Why Me? I can easily do your assistant as I have 5 years of experience in LLM serving, Kubernetes, multilingual NLP, voice pipelines, RAG, and agent orchestration. Not only this, I have a strong grip on security, data privacy and CI/CD. ➡️ Let's have a quick chat to discuss your project in detail and let me show you samples of my previous work. Looking forward to discussing with you in chat. ➡️ Skills & Experience: ✅ LLM Deployment ✅ Multilingual Embeddings ✅ Retrieval‑Augmented Generation ✅ Kubernetes ✅ Docker ✅ Voice Streaming ✅ OCR & PDF Parsing ✅ Agent Frameworks ✅ Security Hardening ✅ Data Privacy (PDPL) ✅ CI/CD Pipelines ✅ Dashboard Development Waiting for your response! Best Regards, Zohaib
$6,000 USD in 2 days
8.0
8.0

Hi, the critical challenge is keeping answers grounded and responsive during peaks, with every model and service hosted inside Saudi Arabia. I’d propose self-hosted Arabic-capable LLMs, hybrid retrieval with multilingual embeddings and reranking, plus structured lookups for fees, deadlines and admissions. Citation checks and abstention would gate answers; “zero fabrication” needs measurable acceptance tests. LangGraph would handle typed, permission-checked tools with retrieval fallback. Local streaming STT/TTS would support barge-in. Ingestion would combine change detection, PDF/OCR parsing, approval and versioned indexes with rollback. Delivery would include the RTL widget, admin portal, Kubernetes/CI/CD, audit logs, PII controls, evaluation and handover. PDPL/NCA requirements would be mapped with your compliance team. Proposed milestones: • Discovery and Arabic/voice benchmark: $3,000. • RAG, ingestion and tools: $9,000. • Voice, widget and admin: $8,000. • Load testing, red-team and handover: $5,000. Estimate: 8-10 weeks; Saudi infrastructure/GPU costs separate. Voice latency and peak capacity need benchmarking. I’m ready for the NDA. Is Saudi GPU infrastructure available, and what share of peak sessions use voice?
$8,500 USD in 26 days
7.1
7.1

Hello, I would be happy to help you build a secure, scalable, bilingual AI assistant for your university, supporting Arabic and English through text and voice interactions. I have experience in AI development, RAG architecture, API integrations, and scalable backend systems. I understand the importance of accurate, source-grounded responses, data privacy, and reliable performance for thousands of concurrent users. I recommend a self-hosted Arabic-capable LLM with hybrid RAG, multilingual embeddings, re-ranking, and hallucination prevention. LangGraph can handle agent orchestration, while FastAPI, Redis, PostgreSQL, and Kubernetes will support scalability. Voice processing will use locally hosted speech models, ensuring all data remains within Saudi Arabia. The solution will include automated content ingestion, source citations, an admin dashboard, security controls, and alignment with Saudi PDPL and NCA requirements. I am comfortable signing an NDA and discussing relevant project references during our initial conversation. Looking forward to working with you. Best regards, Deepak
$8,500 USD in 90 days
7.4
7.4

Hi, I build production RAG, agent, and voice systems (document parsing with OCR, LangChain/LangGraph agents, realtime voice agents, LangSmith observability, PII redaction). Architecture, all self-hosted in a KSA region: - vLLM on Kubernetes serving an open-weight Arabic model (ALLaM, Jais, or Qwen), chosen by Arabic eval scores. - Hybrid retrieval: OpenSearch BM25 + BGE-M3 embeddings, cross-encoder re-ranker, metadata filters, cited sources. - Grounding: claim-level faithfulness check, abstention, and fees/deadlines served from structured data. - Ingestion: incremental crawler, layout-aware PDF/table OCR, versioned index with rollback. - LangGraph agents with typed tools, guardrails, retrieval fallback. - Voice: streaming Whisper STT and Arabic TTS with barge-in. - Admin portal, RBAC, audit log, OWASP LLM controls, eval harness and red-team report. Milestones (12 weeks): 1. Ingestion + grounded RAG + evals: USD 3,500 2. Agents, voice, RTL widget: USD 3,000 3. Admin, security, load test, K8s handover: USD 3,500 To be upfront: my shipped RAG systems have been in English, not Arabic, so I would want a native Arabic speaker from your side in the evaluation loop. References on request. Happy to sign the NDA. Stefan
$10,000 USD in 84 days
6.8
6.8

The hardest part here isn't RAG. It's serving 2,000+ concurrent sessions and sub-2 second voice on self-hosted Arabic models inside Saudi Arabia. If the GPU plan is undersized
$6,750 USD in 30 days
7.5
7.5

Exciting project to work ! Mxpertz Infolabs combines AI chatbot development with full-stack engineering in Java and JavaScript, making us well suited to build a production university assistant rather than a simple website chatbot. Since 2019, our development studio has delivered reliable web and mobile solutions, with experience spanning custom platforms, APIs, integrations and scalable application development. Our AI development capability aligns with the assistant’s conversational requirements, while our JavaScript and Java skills support the responsive RTL widget, administrative interfaces, backend services and integrations. We also understand the importance of clean development, dependable communication and maintainable solutions designed to grow with business needs—key considerations for a 24/7 university service handling substantial traffic. I would welcome the opportunity to contribute as your development partner and deliver a robust, user-friendly AI assistant with complete source-code handover. Best, Meera **Founder
$5,500 USD in 28 days
6.8
6.8

Hey, the detail I'd plan around is how students actually write. Many won't use formal Arabic: they'll type Gulf dialect, Arabizi ("ana abi a3rf el rusoom"), or switch between Arabic and English in one sentence. So the retrieval layer normalizes Arabic text (alef forms, taa marbuta, diacritics) before keyword search, uses multilingual embeddings tested on mixed-language queries, and I'd benchmark candidate models like ALLaM, Jais and Qwen on your own real questions before choosing one. Architecture: vLLM on in-Kingdom GPUs, hybrid BM25 and vector search with re-ranking, and LangGraph agents with typed tools. Fees, deadlines and admission criteria come from structured tables, with the model only phrasing the answer, never computing it. Every claim is checked against its cited source, and the assistant abstains below a confidence threshold. Voice runs on streaming speech-to-text and text-to-speech with barge-in, all self-hosted. Milestones (about 6 months): 1. Ingestion, versioned index and evaluation harness 2. RAG, structured answers and safety layer 3. Agents and tools 4. Voice 5. Admin portal, widget, load test at 2,000 sessions, red-team results and handover Rate: $20 per hour, with each milestone priced after the NDA review. Will the GPUs be provided in your Saudi data center, or should I size and propose the hardware? Juan Pablo
$5,000 USD in 7 days
6.5
6.5

SoftwareLinkers is a verified full-stack and infrastructure team with a 4.9★ rating from 100+ reviews and a 96% job success rate, focused on secure, scalable systems for education and enterprise platforms. That background aligns well with a university assistant requiring reliable backend services, APIs, administration features, monitoring and long-term support. Our skills include AI development, AI chatbot development, AI automation, PHP and AngularJS, enabling us to build the assistant and its surrounding web platform as a maintainable production system. We can support multilingual conversational functionality, structured integrations, role-based administration and scalable data-driven services while keeping security and operational reliability central. Our experience with education and eLearning systems is especially relevant to serving students, applicants, parents, staff and visitors. We understand the importance of Saudi-hosted infrastructure, privacy controls, source-grounded answers and a responsive RTL interface. SoftwareLinkers would be pleased to work under NDA and deliver a well-documented solution with full source-code handover. Choose our team for a dependable technical partner from development through deployment and support.
$8,200 USD in 36 days
6.6
6.6

Hello, I would structure this as a Saudi-hosted, modular AI platform: self-hosted open-weight Arabic-capable models served through vLLM, hybrid BM25/vector retrieval with reranking, and a versioned ingestion pipeline for university web pages and documents. Structured admissions and fee data would handle deterministic answers, while a guarded LangGraph orchestration layer would manage approved tools. Streaming voice, the admin portal, Kubernetes deployment, and evaluation would follow as separate workstreams. I recommend milestone delivery: architecture and infrastructure validation ($1,500, 2 weeks), retrieval and ingestion MVP ($3,000, 3–4 weeks), then voice, administration, security, evaluation, and deployment ($5,500, 5–7 weeks). Full production readiness would require validated Saudi hosting, model capacity, and acceptance testing. Could you confirm whether you already have Saudi-based GPU infrastructure and approved Arabic/English source documents? Have a nice day
$5,000 USD in 60 days
6.1
6.1

Hello, Your load numbers and the rule that nothing leaves Saudi Arabia change the design completely. I would not bolt a cloud chat API onto the site. I would run open-weight Arabic-capable models in-region, with hybrid retrieval, re-ranking, and metadata filters so every answer points back to an official page. Fees, deadlines, and admission rules stay deterministic from structured data, and the system abstains when it is not sure. Proposed architecture in a few lines: incremental crawler with JS rendering and OCR into a versioned index with rollback, LangGraph-style multi-agent layer with typed tools for tickets, contact lookup, and visit booking, streaming voice with barge-in under two seconds, plus an admin portal for review, roles, and audit. Kubernetes, CI/CD, RTL-correct widget, and PDPL/NCA-minded guardrails around the whole path. Timeline: about 30 days for a production-ready first release. Week 1 ingestion and RAG core, week 2 grounding and agents, week 3 voice and admin, week 4 load tests, red-team, docs, and handover. Rate by milestone: architecture and NDA kickoff, RAG plus crawler, agents and safety, voice and admin, then deploy and source handover. Two relevant references: an Arabic campus FAQ assistant with citation-first answers, and a bilingual helpdesk bot...
$8,500 USD in 30 days
5.9
5.9

Hi, Fernando here, from Mexico. Building your university AI with 200,000 queries/month and 20,000 users is doable; what matters most is answers people can trust, at a running cost that stays predictable. I'm a senior AI developer with 12+ years of experience working with JavaScript, AI Chatbot Development, AI Voice Agents and AI Automation. For your university AI, the real challenge is wrong answers and prompt injection. My plan: first, collect real examples and write down the expected result for each. Then, build the LLM pipeline with guardrails and logging. Finally, score it against those examples and tune it before launch. I'd start with the core assistant and its test set as phase one. For the stack, I'd suggest Kubernetes and LangGraph. The price, timeline and milestones are in the bid details. I've done AI work in JavaScript. A few questions to better understand the scope: 1. Do your documents change often? Then automatic re-indexing keeps answers current. 2. Score each change on a fixed set of real questions? I'd recommend it, quality becomes measurable. 3. Who asks the questions, customers or staff? Customers need stricter guardrails than your staff. I can start right away. Happy to go over the details here in chat and suggest the best approach. Looking forward to hearing from you, Fernando
$6,800 USD in 30 days
5.9
5.9

Hello. "ZERO-FABRICATION ARABIC + ENGLISH ANSWERS" — students need reliable university information with official citations, not confident-sounding guesses. I would build this around a Saudi-hosted, open-weight model with hybrid RAG, multilingual embeddings, re-ranking, and source-level grounding. Fees, deadlines, and admission criteria should come from validated structured data, with abstention when evidence is insufficient. The architecture would include incremental content ingestion, versioned indexing, typed agent tools, streaming voice, an RTL-ready admin portal, and Kubernetes deployment. I would also include evaluation, security testing, monitoring, and full source-code handover. The deployment must keep data and inference inside Saudi Arabia, with no external runtime API dependencies. I can define milestones around architecture, core assistant, voice and admin features, testing, and production deployment. Do you already have a Saudi-hosted GPU environment and preferred open-weight model, or should infrastructure and model selection be included in the scope? Looking forward to work with you.
$6,000 USD in 30 days
6.0
6.0

I have built a university chatbot for a Middle Eastern client that handled 150,000 monthly queries in Arabic and English with voice and text, so I understand the scale and bilingual challenges here. For your project, I propose a modular RAG pipeline with self-hosted Arabic-capable embedding models and metadata filters, combined with a deterministic rule-based layer to ensure zero fabrication on structured data like fees and deadlines. Content ingestion will use an incremental JS renderer crawler plus OCR for PDFs and tables, updating the index daily with rollback capability. For voice, I suggest using open-source streaming STT/TTS engines optimized for Arabic, with barge-in to keep responses under 2 seconds. The multi-agent orchestration can be built on LangGraph or a similar open-source framework, integrating ticketing and scheduling tools with guardrails and fallback retrieval. The frontend will be a responsive RTL widget meeting WCAG 2.1 AA, deployed via Kubernetes with full CI/CD pipelines. For safety, I will apply prompt injection filters, PII redaction, and strict scope controls per PDPL requirements. How frequently does your content update, and do you have structured data access endpoints for admission details? Also, do you prefer a specific Arabic embedding model or will I propose one? I can deliver the MVP in 12 weeks with milestones for architecture, ingestion/embedding, dialogue agent, voice integration, admin portal, and final deployment. Ready to sign the NDA and start right away.
$5,000 USD in 7 days
6.0
6.0

The most important architectural decision here is keeping retrieval, inference, voice processing, and operational data entirely within Saudi Arabia while maintaining reliable Arabic and English responses under heavy concurrent traffic. I would approach this with a self-hosted RAG architecture using Arabic-capable open-weight models, hybrid vector and keyword retrieval, a dedicated reranking layer, and claim-level source validation. Admission criteria, fees, and deadlines should come from structured, versioned data rather than generative responses. LangGraph can coordinate specialist agents through typed MCP-style tools, with strict permission boundaries and retrieval-only fallback when tools are unavailable. I would also prioritize incremental ingestion, tenant-aware access controls, prompt-injection defenses, evaluation pipelines, and Kubernetes-based deployment with centralized monitoring and audit logging. My background includes SaaS technical reviews, API integrations, Laravel development, and complex automation involving Make, n8n, Zoho, Shopify, Airtable, and Gmail. I'd be glad to review your architecture under NDA and discuss the implementation approach for your Arabic-first environment.
$5,500 USD in 3 days
6.1
6.1

Building a production conversational assistant for diverse users at your university poses unique challenges, particularly in ensuring all data remains within Saudi Arabia. A robust architecture can utilize self-hosted Arabic-capable models for RAG, ensuring compliance with local regulations. I propose a three-phase approach: 1. Architecture Design: We will establish a scalable framework leveraging Kubernetes for deployment, ensuring compliance and performance. 2. Development: Focus on the RAG question-answering system, utilizing a hybrid retrieval model, and implement the voice pipelines, ensuring low-latency responses. 3. Testing and Deployment: Automated testing for faithfulness and precision, followed by a CI/CD pipeline for smooth updates and a comprehensive admin portal. To align with your hard constraints, we can create a self-hosted solution with no external API dependencies. Could you clarify the specific compliance metrics you wish to prioritize for the Saudi PDPL and NCA?
$5,000 USD in 7 days
5.8
5.8

My background in AI chatbot development and AI development aligns with creating a university-facing conversational assistant, while my experience serving education and technology clients supports clear, accessible user communication. I also offer Arabic and English translation, multilingual transcription, voice-over, technical manuals, FAQs, and privacy-policy writing—useful for a bilingual, voice-enabled university service and its supporting content. My Java, JavaScript, and PHP skills can support chatbot integrations and responsive web experiences, and my 16+ years of professional work with 750+ clients reflect strong documentation, content accuracy, and dependable delivery. However, my profile does not demonstrate the required production experience with self-hosted Arabic RAG, Kubernetes, offline LLM serving, LangGraph/MCP, or Saudi-hosted voice infrastructure, so I cannot accurately claim those capabilities or provide relevant engineering references. I can contribute most effectively to the Arabic-English content, technical documentation, FAQ creation, translation, voice scripts, and conversational copy for this project.
$5,000 USD in 30 days
5.5
5.5

dammam, Saudi Arabia
Payment method verified
Member since May 14, 2012
$10-30 USD
$10-30 USD
$10-30 USD
$10-30 USD
$10-30 USD
$2-8 USD / hour
₹12500-37500 INR
$200-350 USD
$10-30 USD
₹1500-12500 INR
$250-750 USD
$3000-5000 USD
$15-25 CAD / hour
$1500-3000 USD
$30-250 USD
min ₹2500 INR / hour
$100 USD
₹1500-12500 INR
$30-250 USD
$250-750 USD
₹1500-12500 INR
₹1500-12500 INR
$10-30 USD
$30-250 USD
₹100-400 INR / hour