Browse Jobs

We are a technology company developing a mobile outdoor-intelligence system. A sensor unit is mounted on the roof of a suitable vehicle. While the vehicle follows its normal route, cameras, LiDAR, thermal imaging, location data and environmental sensors record the surrounding public space. Our software converts these observations into geolocated detections and signals relating to: * Roads and public infrastructure. * Landscaping and biodiversity. * Traffic and public-space safety. Examples include damaged road surfaces, faded markings, defective streetlights, damaged signs, vegetation blocking visibility, heat stress, unsafe traffic situations and changes that may require further professional inspection. The system is designed as a signalling and decision-support platform. It does not replace engineers, inspectors or asset managers. What already exists We already have the hardware, field datasets, detection results and an existing software platform with a backend and map interface. A detection can contain information such as: * Detection type and ID. * Location and observation date. * Image or sensor evidence. * Confidence and validation status. * Previous observations at the same location. * Route and area information. The platform is still being improved, but the foundation already exists. We are not asking you to build the hardware, detection models or complete platform from scratch. We will provide shortlisted candidates with relevant information about our current backend and software architecture. What we want to add When a user clicks on a detection on the map, a context-aware AI chat should open. The AI must automatically understand which detection has been selected. The user should then be able to discuss the observation, possible causes, relevant projects, possible measures, costs, timing and required verification. For example, the system detects a faded pedestrian crossing. The user asks: Should we repaint this crossing now, apply a temporary measure, or wait and combine the work with the resurfacing project planned for this area? The AI should retrieve the relevant information and explain: * What was detected and how certain the observation is. * Which maintenance plans, projects or logbook entries are relevant. * Which possible measures can be considered. * What acting now may cost. * What may be saved by combining the work with a planned project. * What the possible safety implications of waiting are. * Which information is missing or still requires professional verification. * Which documents and records support the answer. The user must be able to continue the conversation, challenge the suggestion, change the timing and compare alternative measures. Folder-based knowledge structure We do not want the AI to search through an unstructured collection of unrelated files. We expect to connect the chatbot to a logical folder structure containing the organisation’s available knowledge. A possible structure is: Client or Organisation │ ├── Area or Contract │ ├── Multi-year Maintenance Plans │ ├── Planned Projects │ ├── Inspections and Logbooks │ ├── Policies and Technical Standards │ ├── Cost and Unit Rates │ └── Asset Information │ └── Other Areas or Contracts The AI must understand where information comes from within this structure. It should retain and display the complete source context, including the folder path, document title, version, date, page or section. For example: Client A / District 3 / Planned Projects / Resurfacing Programme 2027.pdf / Page 18 Retrieval should be scoped to the relevant client, area, asset, project and time period. Information belonging to another client or area must not be mixed into the answer. Documents should be replaceable or added without manually rebuilding the chatbot each time. The first pilot may use a small sample folder structure. A later phase may connect this structure to OneDrive, SharePoint, cloud storage or another document environment. Phase 1 — the USD 500 pilot For Phase 1, we want one complete and reusable end-to-end use case. The likely demonstration case will be a faded pedestrian crossing. The pilot must deliver: 1. A chat panel that opens from the selected detection in our platform. 2. Automatic transfer of the detection context into the conversation. 3. Ingestion of a small sample folder structure with maintenance plans, projects, logbooks, standards and cost information. 4. Answers with visible and traceable source references. 5. One controlled cost calculation based on rates and formulas supplied by us. 6. A simple comparison between acting now and waiting for planned maintenance. 7. Correct handling of missing, conflicting or uncertain information. 8. Source code, setup instructions and a short handover demonstration. The architecture should be reusable for other detection types after the pilot. A separate generic chatbot or simple “chat with PDFs” demonstration will not be accepted. The chatbot must be connected to the selected field detection and the structured folder context. Important technical principles This phase requires retrieval-augmented generation, or RAG. We are not asking for custom foundation-model training. The AI may retrieve, organise, compare and explain information. It may not invent costs, project dates, policies or safety percentages. Costs must be calculated with structured data, supplied rates and deterministic code. If information is missing, the AI must clearly say that it cannot produce a reliable calculation. Safety consequences must be based on supplied rules, models or documented evidence. If no validated model is available, the system should provide a clearly labelled qualitative comparison instead of inventing an accident probability. The first version must remain a decision-support tool. It must not autonomously create work orders, change asset records or send instructions to external organisations. A human remains responsible for validation and approval. Who we are looking for We need a hands-on AI/RAG integration developer who can work inside an existing software product. You must demonstrate that you have previously built at least one relevant system involving document retrieval, source citations and structured application data. Experience with tool calling, backend integrations, chat interfaces and controlled calculations is strongly preferred. We are interested in what you have actually built. A redacted video, working demonstration, repository, architecture example or client reference is acceptable. Clearly explain which parts you personally implemented. Generic chatbot templates, copied proposals and theoretical AI knowledge are insufficient. You must be willing to: * Work with our existing codebase. * Deliver the completed source code to our repository. * Explain your technical choices. * Provide setup and handover documentation. * Sign an NDA if required. * Transfer the agreed intellectual property upon completion and payment. How to apply Start your proposal with: FIELD-AI-500 Then answer these five questions: 1. Do you accept the fixed total Phase 1 budget of USD 500? 2. Show one comparable RAG or AI-copilot solution you personally built. 3. How would you combine a selected detection with a folder-based knowledge structure? 4. How would you ensure that sources, costs and safety conclusions are not fabricated? 5. What do you need from our backend team, and how quickly can you deliver the pilot? Applications above USD 500 or without a relevant previous example will be rejected. Opportunity after the pilot A successful pilot may lead to a substantially larger assignment involving additional detection categories, live document synchronisation, asset-system integrations, advanced scenario calculations, multilingual interaction, approval workflows and deployment for multiple organizations. For now, the assignment is simple and focused: Build one working, source-backed AI decision copilot inside our existing platform.

Test 02 - QA-only test job for verifying the new hourly bill-after-period billing flow (weekly invoice submission, off-session charge, 14-day hold). Not real work.

Create a HR app for a meduim sized comapny with all the benefits and pay related details.

it burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hell

asdfasdait burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hell What I need help with: asit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hellit burns credit in hell

AI-powered coaching app that helps users interact with a personalized AI mentor. The system provides ongoing guidance, goal tracking, and tailored recommendations based on user behavior and history. We’re looking for a full-stack developer or team who can help build the MVP and core AI experience, not just a chatbot, but a structured AI coach with memory and personalization. What You'll Do Build AI chat and voice-based coaching features using LLM APIs (OpenAI / Claude) Implement user personalization, memory, and context handling Develop onboarding flows to capture user goals and preferences Create backend services for AI orchestration and user management Design scalable API architecture for mobile app integration Work on knowledge ingestion and retrieval for coaching content Tech Stack (Preferred) React Native or Flutter Node.js, Java Spring Boot, or Python FastAPI PostgreSQL, Redis OpenAI / Anthropic APIs AWS or GCP Requirements Strong full-stack development skills Experience with LLM integration and conversational systems Understanding of AI memory, RAG, or personalization systems Mobile development experience (iOS/Android) Nice to Have Experience with AI agents or workflow systems Voice AI / speech interfaces Background in coaching, education, or productivity apps Deliverables Working AI coaching MVP (mobile app + backend) User memory and personalization system Documentation for future expansion