Expert AI/LLM Engineer Needed – SVG Template Automation, Accuracy & Cost Optimisation (Paid Sandbox)

# Expert AI/LLM Engineer Needed – SVG Template Automation, Accuracy & Cost Optimisation (Paid Sandbox) **Experience Level:** Expert Only **Project Type:** Paid AI proof of concept + potential technical consultation **Budget:** [TO BE CONFIRMED] **Existing Tech Stack:** Not required for this independent sandbox project. ## Project Overview I'm the founder of a SaaS platform that has been under development since April 2026. Approximately **80% of the platform is complete**, with most functionality already developed and tested. However, we've encountered challenges with **AI-powered SVG template selection and automatic content population**, particularly around visual accuracy and excessive AI API costs. Our latest internal tests achieved an estimated 65–70% accuracy, but we need a significantly more reliable and cost-effective solution. I'm looking for a **highly experienced AI/LLM specialist** to independently demonstrate a working solution in a sandbox environment before potentially consulting with my existing development team. **No access to our existing codebase, infrastructure or production environment will be required.** ## 1. The Technical Challenge We have professionally designed SVG templates, each with predefined Canva annotations and JSON rules specifying editable versus locked elements. The required workflow is: 1. **Input analysis:** Analyse a supplied text-based input. 2. **Template selection:** Evaluate approximately 10 predefined SVG templates and select the most suitable one. 3. **Template population:** Automatically populate the selected template with relevant text and permitted colour adjustments, following the predefined editing rules. 4. **Layout preservation:** Maintain the original design, including locked elements, correct text positioning, wrapping, spacing and formatting. 5. **Output:** Produce a clean, professional visual suitable for editing and export. ### Current Problems - Inconsistent template population and visual accuracy. - Editable fields occasionally left blank. - Text wrapping, positioning and layout issues. - Earlier tests produced corrupted symbols, although this appears resolved. - Excessive AI API costs. **Two critical requirements:** - **Accuracy:** Consistently high-quality results across multiple templates and inputs, without missing content, overflow or unintended layout changes. - **Cost efficiency:** Minimise AI calls, tokens and retries while maintaining quality. Costs must be measurable and commercially sustainable. ## 2. Paid Sandbox Project Following an initial qualification meeting, I will provide: - Approximately 10 SVG templates. - Corresponding Canva annotations and available JSON rules. - Sample inputs and expected behaviour. - An agreed AI API testing budget, covered separately from your development fee. Your responsibility is to build an **independent, working sandbox** demonstrating the complete workflow. Deliverables must include: - Working template selection and automatic population. - Correct handling of locked/editable elements. - Consistent visual rendering across repeated tests. - Accuracy and reliability results across varied inputs. - Transparent API cost reporting per successful generation, including retries. - Sandbox source code, configuration and documentation. You may recommend whichever AI model, API and technical approach is most suitable, provided it is reliable, cost-effective and independent of our existing codebase. ## 3. Three-Stage Hiring Process **Stage 1 – Initial Qualification Meeting** I will explain the product, current issues and requirements to shortlisted candidates. I expect you to demonstrate relevant experience and outline a credible technical approach. I'm not looking for someone learning this technology from scratch. **Stage 2 – Paid Sandbox Development** If shortlisted successfully, you will be hired for a clearly scoped, paid sandbox project. I will cover the agreed AI API costs separately. The goal is to demonstrate high accuracy, reliable outputs and cost efficiency across multiple tests. **Stage 3 – Paid Consultation With My Existing Development Team** If the sandbox is successful, I will arrange a separate **paid 1–2 hour technical meeting** with my existing developers. You will explain your solution, demonstrate the sandbox, discuss potential causes of our current issues, and advise how your approach could be integrated into the existing SaaS. Any further development assistance would be agreed separately. ## 4. Required Expertise Strong, practical experience with: - LLM APIs, structured JSON outputs and AI workflow architecture. - SVG manipulation and dynamic template population. - Automated text fitting, wrapping, layout preservation and rendering. - AI model selection, token optimisation and API cost reduction. - Building reproducible technical proofs of concept. **Previous experience building comparable AI-powered visual generation systems is strongly preferred.** I'm specifically seeking an expert, not a general developer whose experience is limited to basic chatbot or LLM integrations. ## 5. Expectations & Application I expect a **clear technical strategy, proven experience and an emphasis on reliability and cost effectiveness**. I'm not interested in open-ended experimentation or funding someone to learn the fundamentals. Necessary technical validation is expected, but it should follow a structured plan with measurable results. When applying, please briefly provide: 1. Examples of similar AI/SVG generation systems you've worked on, including your specific role. 2. Your proposed high-level approach to this challenge. 3. Your experience optimising AI API costs. 4. Your availability and estimated sandbox fee. 5. Your hourly rate for the potential follow-up consultation. **Please only apply if you're highly experienced and confident in tackling this specific type of AI-powered visual template system.** The objective is straightforward: **prove that this workflow can generate consistently professional SVG visuals at a commercially viable cost, then help our existing team understand how to implement it.**