Natürliche Sprachverarbeitung Jobs

I want to replace the manual US Calling sales calls my team is making with a fully automated, AI-driven outbound dialer. The system has to plug straight into our existing CRM so that every call is launched from, and written back to, a single customer record. Core workflow • The dialer should pull a lead from the CRM, place the call automatically, open a tailored script, and capture everything the prospect says through real-time speech recognition. • Natural language processing must understand intent and sentiment, while AI-driven responses keep the conversation moving until a meeting is booked or a hand-off to a human rep is required. • Once the call ends, the transcript, outcome, and next-step task should save to the same CRM record without anyone touching a keyboard...

Hi — we’re looking for an experienced AI engineer who has built real-time conversational voice systems. Our platform uses AI to answer phone calls for businesses, take orders, and interact naturally with customers. The system integrates voice AI, APIs, and a backend dashboard. We’re looking for someone who has experience with things like: • real-time voice AI • conversational agents • speech-to-text / text-to-speech • API integrations • production systems that handle live calls This would initially be for reviewing and improving an existing system, with the possibility of ongoing work if it’s a good fit. If you have experience building voice AI systems, please send examples of similar work you’ve done. Thanks.

We are looking for a web designer to create a corporate website (14 pages) for a company specializing in construction material installation and public facility construction. Requirements: Responsive design (PC / Tablet / Mobile) Clean and modern design Attention to detail -We Provide: Text and images Preference: Female designers preferred. Please send: Portfolio Estimated timeline Your price. reference site:

I want a small-footprint AI integration that watches every new email that lands in my inbox and each incoming WhatsApp message, immediately scans the content for two kinds of keywords—my product names/services and a few predefined phrases or topics—and then pushes a concise summary back to me on both channels. The flow I imagine is straightforward: • Trigger: new email or WhatsApp message • Action: keyword detection + short, plain-language summary (no auto-reply needed at this stage) • Delivery: the summary is sent to my email and appears in my WhatsApp chat so I never miss an important lead or topic. Key points you should know • Flexibility to expand or edit the keyword list without touching core code. • Accuracy matters more than speed; a 30-60 ...

I want to launch a cloud-based lending platform similar to CredOrbit, but fully white-label and multi-tenant so Chartered Accountants (CAs), Direct Selling Agents (DSAs), end-customers, and partnered banks can each log in to their own workspace. The core engine must be AI-driven: it should analyse borrower data, flag risk, pre-fill forms, and surface next-best actions for every user type. The site itself will operate on a classic SaaS model—users subscribe, create their organisation space, and start originating or reviewing loans. Please refer the overall flow and feature depth you see on , similar solutions to be provided. Key functions I need you to deliver • AI underwriting & insights – Pull KYC, bank statements, GST and credit-bureau data, then run rule-ba...

I’m building a distributed network where natural-language AI agents can discover each other, exchange insights, and tap into spare GPU capacity offered by compute providers. The immediate target is a functional MVP that clearly demonstrates enhanced AI-agent collaboration—our primary goal—within one month. What’s needed now is a peer-to-peer coordination layer that registers agents and compute nodes, matches them in real time, and records usage on-chain. Smart contracts (EVM-compatible) will govern identity, staking, and reward distribution, while a REST or GraphQL gateway plus a lightweight SDK will let developers plug their NLP models into the network with minimal effort. Docker-based reference nodes and clear step-by-step documentation should make it simple for ...

I need an end-to-end solution that lets inspectors interact with an AI assistant over both voice calls and WhatsApp text. While they are on-site the assistant should prompt them, capture their spoken or typed notes, and request photos where relevant. All of that input—images plus text notes—must be automatically organised by product and compiled into a clean, branded PDF report that can be emailed or downloaded the moment the inspection is finished. Key expectations • The AI handles both channels seamlessly: a quick call for walk-through narration or a WhatsApp chat for typed updates and photo uploads. • Image and note metadata are tagged to the appropriate product so the final document is easy to review. • The PDF is generated in a professional layout w...

I need an end-to-end AI agent that can sit inside my WhatsApp Business account and handle the full sales cycle for bulk Polyethylene (PE). The bot must be able to open a chat proactively, answer every follow-up question in real time, negotiate prices, payment terms and delivery schedules, and finally secure written confirmation inside the same thread—all while sounding friendly and conversational, just like my own voice. You’ll receive a large, well-labelled archive of past WhatsApp conversations that already closed real PE deals. Use that data to train or fine-tune your model (LLM, RAG, or any hybrid approach you prefer) so the agent learns our typical pricing logic, MOQ limits, freight clauses, and cultural nuances. Once live, it should keep learning from fresh chats I appro...

I’m building a fully-featured voice assistant that users can download on both iOS and Android. Think of it as a pocket-sized helper that can listen, understand, and act on spoken commands—setting reminders, fetching quick answers, controlling smart-home services via API, or triggering in-app actions. Here’s what I need from you: • A native or cross-platform mobile app (Swift/SwiftUI + Kotlin, or Flutter / React Native—whichever lets us land on both stores quickly and cleanly). • On-device and cloud speech recognition so the assistant responds within a second or two, even on slow connections. • Natural-language understanding: basic intent detection, entity extraction, and a fallback to a large-language-model endpoint when the user asks something ...

I want an AI-powered conversational agent that has to work on a web application front end. The bot must tap into our large retail product database so managers can conversationally, ask for actionable insights, kpi, and create reports. Key goals • Deliver actionable insights about consumer behaviour • Easy conversational chat for business user Acceptance criteria documentation covering model architecture, training data structure, and how to connect to the retail database API. 3. Handover of all code (Python/Node or comparable), configuration files, and instructions so my in-house devs can maintain it. 4. Analytics dashboard reporting 5. Final deployment on our production site with my sign-off after load testing. If you have experience with AI projects using LangGrap...

Our labelling process consist of accurate segmentation of the sound wave of audios, selection of the role and related attributes of speaker according to their voice, and transcription of the speech through our TMS (Transcription Management System) platform. Please share your previous experience with German text and audio annotation, if applicable. If you have any questions, feel free to message us on Freelancer. We look forward to receiving your application. Thank you.

I want to create an AI-driven application that translates seamlessly between Amharic and English. The core of the product is high-quality language translation, but it must also include a full voice-recognition layer: • Speech-to-text conversion so users can dictate Amharic or English and see accurate transcripts. • Voice-command execution to let them control key in-app actions hands-free. • Language identification that automatically detects whether the incoming audio or text is Amharic or English before processing. The finished app should run smoothly on Android phones and as a responsive web application; a shared code-base or robust API that feeds both fronts is ideal. I’m open to your preferred stack—TensorFlow, PyTorch, Whisper, or any other modern NLP...

I need an outbound voice AI that can pick up the phone, speak naturally with customers, and lock in service appointments without human involvement. The bot should dial leads from a list, confirm the customer’s need for our service, offer available slots pulled from an online calendar, secure the booking, and then push the confirmed appointment back to our CRM while sending the customer a confirmation SMS or email. Core flow • Call initiation through a voice gateway such as Twilio • Speech recognition and NLU (Dialogflow CX, Amazon Lex, Rasa, or similar) to understand the caller’s intent, handle clarifying questions, and gracefully manage “I didn’t catch that” moments. • Text-to-speech that sounds human; Google WaveNet, Amazon Polly or Micro...

I have an Arabic article about Arabic coffee (قهوة عربية) – its cultural significance, preparation rituals, and role in hospitality across the Gulf and Arab world. The article covers: The history of coffee in the Arab world How Arabic coffee is different from other coffees (lighter roast, cardamom, saffron) The dallah (coffee pot) and finjan (small cup) – their designs and meanings The ritual of serving – who gets served first, how to accept, how to refuse Coffee and dates – why they're always served together Regional differences (Saudi, Emirati, Omani, Jordanian) Modern coffee culture vs. traditional majlis coffee UNESCO recognition of Arabic coffee as intangible cultural heritage What I Need I need the full article translated into clear, natural Engl...

Build an end-to-end AI application that can reliably solve JEE-style math problems, explain solutions step-by-step, and improve over time. The goal of this assignment is not just model usage, but to evaluate whether you can: design a RAG pipeline build a multi-agent system handle image, text, and audio inputs introduce human-in-the-loop (HITL) implement memory & self-learning package everything into a working application and deploy it You do not need to be a DevOps or full-stack expert, but you must be able to build, run, and deploy a simple app.

I need an AI-powered chatbot fully integrated with the WhatsApp Business API. It must converse fluently via text, understand incoming voice notes, and react appropriately to images or short video clips sent by users. I’m open on the underlying stack—Dialogflow, Microsoft Bot Framework, IBM Watson, or any other platform you believe best fits WhatsApp’s constraints—so long as latency stays low and the solution can scale as traffic grows. Core deliverables: • End-to-end WhatsApp Business API setup (webhook, number verification, cloud or on-prem hosting). • NLP pipeline that handles:  – Text intent recognition and response generation.  – Speech-to-text for voice messages, with the transcript feeding the same intent flow. &...

I’m building a voice-driven receptionist that greets visitors or callers, listens to their questions, and replies with a natural-sounding voice. The core flow is straightforward: • Speech-to-text: Accurate recognition of English only is required for now. • NLP: Classify the intent and pull the correct answer when a caller asks about our business hours or location. • Text-to-speech: Respond with friendly, human-like audio. • Fallback: whenever the system is unsure, it should politely ask for more details rather than handing the call off or giving a canned line. I need the full stack—STT, intent handling, response generation, and audio playback—wrapped in a module that I can drop into my existing website widget today and expand to a phone line ...

I want an offline-running AI that studies each WhatsApp conversation, decides if a follow-up is actually necessary, chooses the best moment to send it, then drafts and fires off a friendly, contextual message—all without keyword tricks. The single most important benefit for me is saving the hours my team now spends chasing silent leads, so the system must be accurate enough that we can trust it to work on its own. How I picture the flow 1. A new or ongoing chat lands in our WhatsApp Business account. 2. Your model (hosted locally, not in the cloud) pulls the full thread, understands intent, sentiment, and stage of the sale from overall context. 3. If follow-up is warranted, it schedules the outreach intelligently—e.g., after an unanswered quote give it six business days,...

I’m building a voice-driven receptionist that greets visitors or callers, listens to their questions, and replies with a natural-sounding voice. The core flow is straightforward: • Speech-to-text: Accurate recognition of English only is required for now. • NLP: Classify the intent and pull the correct answer when a caller asks about our business hours or location. • Text-to-speech: Respond with friendly, human-like audio. • Fallback: whenever the system is unsure, it should politely ask for more details rather than handing the call off or giving a canned line. I need the full stack—STT, intent handling, response generation, and audio playback—wrapped in a module that I can drop into my existing website widget today and expand to a phone line ...

I want to roll out an AI receptionist whose single role is to speak with customers and give them clear, accurate information about our products and services. A caller (or site visitor) asks a question, the system turns that speech into text, pulls the correct answer from a knowledge base I will supply, and then replies in smooth, human-sounding audio—no appointment booking, no generic FAQ chatter, just concise product guidance. Here is the scope I have in mind: • Speech-to-text captures the question. • An NLP layer (LLM, Dialogflow, Rasa, or a framework you recommend) matches the query to the right answer. • Text-to-speech returns the response in natural audio, ideally with a custom voice so it feels branded. • The module plugs into either my website chat ...

I want an end-to-end Claude-powered agent that brings all of my buy-side research workflows under one roof. The agent must • Automatically pull PowerPoints, “Concall Transcripts”, annual reports from BSE or , and broker research (updates, downgrades, initiating-coverage notes) straight out of my one-year email archive • Parse each file, detect the company it belongs to, and save it in a dedicated Drive folder for that company—no mix-ups • Pass every saved document into NotebookLM and spin up a fresh notebook per company, ready for question-answering inside NotebookLM • Sit inside Microsoft Teams as a bot so I can type, for example, “Create notebook for HDFCBANK” or “What did the latest downgrade say about margins?” and get...

Our Intercom inbox relies on FIN AI, yet too many chats still reach human agents. After digging into the analytics I’m convinced the root cause is the accuracy of the answers FIN serves, not the UI or core algorithm. I need an expert who can turn better training data into sharper, more dependable responses so customers self-serve more often. Here’s what I’m looking for you to do: • Audit recent conversation logs and the existing FIN knowledge base to pinpoint coverage gaps and misleading examples. • Curate or create high-quality, well-structured training data that directly addresses those gaps—think clarified FAQs, edge-case scenarios, and fresh intents drawn from real chats. • Prepare the dataset in the format Intercom accepts and guide (or h...

I need a ChatGPT-style assistant that can read every file my organisation generates—PDFs, Word docs, Excel sheets, HTML pages and anything else we store—then answer questions using that private knowledge base. The assistant must recognise the language of the query and reply in the same tongue; Hindi, Marathi, English and Telugu are the priority. The job covers the full pipeline: ingesting and parsing documents, building a semantic index or vector store, connecting it to a generative model and exposing everything through a simple chat interface. Whichever mix of OpenAI, open-source LLMs, embeddings or retrieval-augmented generation you choose is up to you, as long as the final experience feels as fluid as ChatGPT while keeping my data secure on our own servers. Please send a d...

I’ve built a hallucination-resistant LLM platform that already marries FAISS-based retrieval, answer cross-checking and modular knowledge-base support. The last piece that still feels brittle is the lightweight fine-tuning component. Today it performs well on the narrow domain data I trained on, yet I need it to generalize confidently across both technical subjects and varied industry-specific information without bloating compute costs. Here’s what I’m after: • Refresh the current LoRA/PEFT workflow or suggest an alternative that can stretch the model’s knowledge boundaries while keeping the footprint “lightweight.” • Curate or synthesize balanced tuning and evaluation sets that cover the two priority areas (deep technical content and sect...

Real-Time Face Swap System Requirements Specification 1. Project Overview Purpose: Development of real-time face replacement functionality aimed at protecting privacy and expanding expressive capabilities during streaming. Users: Exclusively for individual streamers. Streaming Environment: - Current: Mac mini (M4 chip, 16GB RAM) - Future: Upgradeable to 32GB/64GB as needed - OS: Mac preferred (portability priority), Windows compatibility considered if necessary - Streaming Software: OBS to be used 2. Non-Negotiable Requirements (Must Have) 2-1. Technical Requirements Face Replacement Specifications: - Replace only eyes and eyebrows - Use original parts for nose down (mouth/lips) - Reason: To prevent double lips when drinking liquids Safety: - Face processing must never drop out w...

Dazzx is moving beyond simple template filtering toward a genuine Creative Layout Intelligence Engine. The platform’s canvas already renders static JSON designs; what is missing is the brain that, from three user inputs—uploaded assets, a short marketing prompt, and the chosen design size—can suggest the three strongest advertisement layouts out of a library of 10,000-plus templates. The assignment is to build that brain. The workflow I have in mind is: • Parse and understand the marketing prompt, extracting intent, tone, and any product cues. • Match those insights against our tagged template corpus, scoring each layout on visual hierarchy, probable asset fit, and relevance to the prompt. • Return the top three candidates as Fabric-compatible JSON, ...

I want to launch a local delivery service here in Italy under my company name, Bistrosoftware. To do that I need a small web-based ordering system built around a chat interface that works equally well in text and voice. Here’s the flow I have in mind. A customer opens the widget on our site (or a standalone landing page), chooses either Italian or English, speaks or types an order, and the AI takes it from there. The bot must: • understand natural language questions about the menu (advanced NLP) • guide the customer through item selection, quantities, address and preferred delivery time (basic response handling) • present the PayPal checkout, confirm payment, then return an order number • keep the customer updated through the same chat thread as the ord...

I need an experienced AI/ML freelancer to take the lead on building a custom copilot that streamlines day-to-day clinical and administrative workflows for a healthcare platform. The immediate goal is to move from scattered manual steps to an intelligent assistant that can understand natural-language requests, pull data from our EHR, draft documentation, route approvals, and trigger follow-on actions inside existing tools. You will have full freedom to choose the right blend of large language models, retrieval-augmented generation, and automation frameworks, provided the result is HIPAA-compliant, auditable, and easy for my in-house dev team to maintain. LangChain, FastAPI, vector stores such as Pinecone or FAISS, and orchestration services like Airflow or Prefect are all on the table if t...

Infinity is the name I have reserved for an AI-driven web platform that functions first and foremost as an educational resource. I want a clean, scalable solution that lets anyone—from K-12 pupils to college, postgraduate or lifelong learners—store, organise, and retrieve knowledge in any field with the help of machine-learning features. Here is the vision in plain terms. Users create an account, upload or link their study material (PDFs, videos, lecture notes, articles or datasets), and the system instantly classifies, tags and indexes that content. A conversational AI layer (think ChatGPT-style) then allows them to ask questions, generate quick summaries, build flashcards, or create personalised study paths from their own resources plus open-access reference material pulled ...

My goal is to build an AI-driven pipeline that can read PDFs of medical subjects packed with both text and images, understand the material, and turn every chapter into student-friendly assessments. For each file the system should automatically deliver short-answer questions that capture key facts, long-answer prompts encouraging deeper explanations, and well-structured multiple-choice questions—complete with one correct option and plausible distractors. Because many of the PDFs include labelled diagrams, charts, or photographs, the engine must combine standard text extraction with reliable OCR so information buried in images feeds directly into at least one generated question. The tone and vocabulary need to suit secondary-school students, keeping accuracy on scientific terminology ...

I’m looking to have a fully-functional AI receptionist built that can pick up every inbound call, hold a natural conversation with the caller, and then route the call to the right person or department purely on the basis of what the caller actually says. The key outcomes I need are: • Advanced conversational AI – the system must go well beyond simple keyword spotting and understand context, intent, and follow-up questions so the interaction feels human. • Dynamic query-based routing – once the caller’s need is clear, the AI should transfer the call to the appropriate extension or external number automatically. • Clean hand-off – when the call is routed, the receiving party must get a short, accurate summary of the caller’s request...

I need a purpose-built CRM for the strata industry that leans heavily on AI to wipe out repetitive admin and give committees, owners and managers instant insight. The solution must weave in predictive analytics to flag maintenance and arrears before they surface, natural language processing so residents can lodge requests or search by-laws in plain English, and automated data entry that lifts information straight from invoices, minutes and inspection reports. The underlying platform is your call—web, mobile, desktop or a hybrid—so long as it comfortably scales to thousands of lots and high concurrent traffic without performance dips. Core functional areas I have in mind: • Centralised contact & lot database with role-based access • Ticketing thread that auto-cl...

We’re looking for annotators to work on text-based tasks focused on validating accuracy, tone, and reasoning. The role involves editing and refining content, as well as creating structured reasoning chains that apply logical and linguistic consistency to complex customer–agent scenarios. Candidates should have a Bachelor’s degree, strong fluency in written English, and familiarity with composing or refining AI prompts. We’re looking for individuals with solid analytical and deductive thinking skills, an understanding of online shopping terminology, and strong attention to detail. The work will focus on three main areas: 1. Content Writing & Refinement Edit and refine AI-generated e-commerce content to ensure it sounds natural, authentic, and aligned wi...

Multilingual Voice Recording Project – Code-Switching Conversations Project Name: BV Project Type: Remote | Ongoing (Limited Slots per Locale) Project Overview Project BV is a multilingual speech data collection initiative designed to enhance Automatic Speech Recognition (ASR) systems for high-value multilingual call center scenarios, including financial services, healthcare, and telecommunications. The project focuses on collecting natural code-switching conversational audio, where speakers alternate between two languages within a single conversational turn. Scripts will be provided; however, natural delivery, fluency, and context-appropriate language switching are essential. Language Requirements Primary Language (Native level – one required): Arabic (arAE:UAE...