Optimize AI Interview Platform (Gemini API)

Fix AI Interview System (Gemini API + TypeScript) | LLM Optimization & Backend Stability Project Description We are building an AI-powered interview platform that conducts structured interviews, evaluates candidate responses, and provides scoring using the Gemini API. Currently, we are facing multiple issues with AI response quality, consistency, and system reliability, and we are looking for an experienced LLM / Generative AI Engineer to fix and optimize the system. Current Problems Inconsistent AI responses during interviews Poor structured output (JSON breaking / invalid formats) Weak evaluation logic (incorrect scoring / feedback) Prompt instability across sessions Latency issues and delayed responses Occasional irrelevant or hallucinated answers Difficulty maintaining context in multi-step interviews Scope of Work The freelancer will:

1. Prompt Engineering & AI Optimization Improve prompts for consistent, accurate responses Design structured output formats (strict JSON responses) Reduce hallucinations and irrelevant outputs Implement prompt chaining / system prompts
2. AI Evaluation System Improvement Fix scoring logic for candidate answers Improve evaluation consistency Ensure reliable feedback generation
3. Gemini API Integration Fixes Debug and optimize Gemini API usage Handle rate limits, retries, and failures Improve response time and streaming (if applicable)
4. Backend Stability (TypeScript / Node.js) Fix API orchestration issues Improve session/context handling Ensure smooth multi-step interview flow
5. Guardrails & Validation Enforce structured outputs (JSON schema validation) Add fallback and retry mechanisms Improve overall system reliability Tech Stack TypeScript / Node.js Gemini API (Google AI) REST APIs / possibly WebSockets JSON-based evaluation system *Requirements (MUST HAVE) Strong experience with LLMs (Gemini / OpenAI / Claude) Proven work in AI-based applications (chatbots, evaluation systems, etc.) Expertise in prompt engineering & structured outputs Experience with Node.js / TypeScript backend systems Understanding of AI limitations, hallucinations, and guardrails Nice to Have Experience building AI interview / assessment platforms Knowledge of RAG, embeddings, or evaluation frameworks Experience with real-time systems Deliverables Stable and consistent AI interview responses Structured and valid JSON outputs Improved evaluation accuracy Reduced latency and failures Clean, maintainable code improvements Project Type Short-term (initial fix: 1–2 weeks) Possibility of long-term collaboration To Apply, Answer This (Important — this filters bad freelancers) Have you worked with Gemini or other LLM APIs? Explain briefly How do you ensure structured JSON output from LLMs? How do you reduce hallucinations in AI responses? Share a similar project you’ve worked on Budget: INR 1500–12500 Skills: Node.js, JSON, Typescript, Backend Development, API Integration, REST API, Prompt Engineering, Generative AI, AI Model Development