The Role

The LLM Architect will own Droidal's large language model strategy and architecture for RCM AI agent development. You'll design the AI reasoning layers that allow Droidal's agents to understand payer rules, interpret EOBs and remittance advice, classify denial reasons, generate prior authorization narratives, and make workflow decisions all in a production healthcare environment where accuracy and auditability are non-negotiable.

This is not a research role. You'll be building production LLM systems that process real claims, make real decisions, and directly impact provider revenue.

What You'll Do

Architecture & Design

- Define Droidal's LLM system architecture: model selection, fine-tuning strategy, RAG pipelines, prompt engineering standards, and agentic reasoning frameworks
- Design multi-agent LLM architectures where agents collaborate to complete multi-step RCM workflows (e.g., eligibility authorization submission denial handling)
- Architect retrieval-augmented generation (RAG) systems over RCM-specific knowledge bases: payer policies, CPT/ICD coding rules, prior authorization requirements, and denial categorization logic
- Design prompt engineering standards and guardrails that ensure consistent, auditable, and hallucination-resistant outputs in clinical and billing contexts

Production AI Systems

- Build and own the LLM inference and orchestration layer — integrating with Droidal's AgentFlow platform and backend Python services
- Establish evaluation frameworks: automated evals, human review pipelines, and regression testing for LLM-powered workflows
- Implement observability, logging, and audit trails for all LLM decisions — critical for HIPAA compliance and customer trust
- Define latency, cost,



and accuracy trade-off frameworks for model selection and serving strategy

Voice AI
- Architect the language understanding and generation components of Droidal's Voice AI agents — real-time speech-to-text, intent classification, dialogue management, and payer IVR navigation
- Design adaptive conversation flows that handle the variability of payer IVR systems and patient interactions

Cross-Functional Leadership

- Partner with Staff and Senior Staff Backend Engineers to define the integration surface between LLM layers and production services
- Work with the deployment team to ensure LLM components meet Droidal's security model: data isolation, VDI constraints, customer VPC architecture, and zero-data-exfiltration requirements
- Stay ahead of the LLM landscape — evaluate new models, frameworks, and techniques and translate them into actionable roadmap recommendations
- Mentor ML engineers and provide technical guidance on LLM best practices across the org

What You Bring

- 8+ years in ML/AI engineering, with 2+ years focused specifically on large language models in production environments
- Deep, hands-on expertise with LLM orchestration frameworks: LangChain, LlamaIndex, LangGraph, CrewAI, or equivalent
- Solid experience building RAG systems: chunking strategies, embedding models,



vector stores (Pinecone, Weaviate, pgvector, etc.), and retrieval evaluation
- Proficiency in Python; comfort working alongside backend engineers on system integration
- Experience with both open-source (LLaMA, Mistral, etc.) and proprietary (OpenAI, Anthropic, Google) LLM APIs — and the trade-offs between them
- Track record of building agentic systems: tool use, function calling, multi-step reasoning, and agent evaluation
- Strong understanding of LLM failure modes: hallucination, context window limitations, prompt injection, and output inconsistency — and mitigations for each
- Experience operating AI systems in regulated industries (healthcare, finance, legal) is a strong plus

Nice to Have

- Healthcare / RCM domain knowledge: familiarity with claims workflows, EOB interpretation, denial codes, payer rules
- Experience with speech/voice AI: ASR, TTS, dialogue systems, or real-time conversation pipelines
- Fine-tuning experience: LoRA, PEFT, instruction tuning on domain-specific datasets
- Published work, open-source contributions, or conference talks in the LLM/NLP space

Why Droidal

- Define LLM architecture from the ground up for a platform already at production scale in healthcare
- Your work will directly shape how 2,100+ provider locations recover revenue and reduce administrative overhead
- Tight feedback loops — Droidal's 8-step customer journey gets agents live in days, so you see the impact of your work quickly
- Mission-driven, technically ambitious team; remote-first; high autonomy

*Droidal is an equal opportunity employer. We are committed to building a diverse and inclusive team.*

📌 LLM Architect (Tamil Nadu)
🏢 Droidal
📍 Tamil Nadu