AI Program and CRM Completion

AI Voice Assistant Developer Needed – Fix & Improve Existing CRM-Integrated AI System (Law Firm) Project Overview My law firm has a custom-built, voice-controlled AI assistant that integrates with our CRM — think "Jarvis from Iron Man" for legal case management, document drafting, and web research. I can speak natural-language requests to it, and it interprets the command and executes actions directly in our CRM (pulling records, updating case data, etc.). The system was built in-house but is currently unreliable — it misinterprets commands, occasionally executes the wrong action, and generally needs a more experienced hand to stabilize it. I'm looking for an experienced AI engineer to diagnose and fix the existing issues, and then continue as an ongoing freelance developer to further build out the platform. This is NOT a general "full-stack developer" job — I need someone with specific experience building reliable, production-grade LLM agent systems that take real actions via API calls. What I Need Help With Phase 1 – Diagnostic & Fix Review the existing codebase and architecture Identify why the system produces inconsistent/incorrect results (prompt design, lack of guardrails, poor error handling, API integration bugs, etc.) Fix reliability issues so commands are interpreted and executed correctly and consistently Phase 2 – Ongoing Development Improve error handling, logging, and testing so issues can be caught before they affect real client data Extend functionality as our needs grow Ongoing maintenance and ad hoc support Ideal Skills & Experience Proven experience building LLM-powered agentic systems with function calling / tool use (not just chatbots — systems that take real actions via APIs) Strong API integration experience, ideally with CRM platforms Experience with speech-to-text / text-to-speech pipelines (e.g., Whisper, ElevenLabs) if voice input/output is involved Solid understanding of prompt engineering and LLM reliability practices — building in guardrails, validation, and fallback behavior, not just prompt tweaking Comfortable working with sensitive/confidential data (this is a law firm — confidentiality is critical) Strong communicator who can explain technical tradeoffs in plain language.