Placeholder visual — project screenshots coming soon
Overview
An AI-first “Log HCP Interaction” screen for pharma field reps. The left panel is a read-only form — it is never filled manually. All data entry happens through the AI assistant chat on the right, powered by a LangGraph agent calling Groq’s gemma2-9b-it.
Built for a technical assessment: design an AI-powered CRM tool for Healthcare Professional (HCP) interaction logging, using LangGraph and an LLM as mandatory components.
Features
- (001)log_interactionExtracts HCP name, date/time, topics, sentiment and materials/samples from free text and fills the form
- (002)edit_interactionPatches only the fields the rep says are wrong
- (003)suggest_followupThe LLM proposes 3 concrete next actions based on the logged interaction
- (004)detect_adverse_eventPharmacovigilance-style check for patient adverse reactions mentioned in the interaction
- (005)search_past_interactionsAnswers natural-language questions about a HCP’s history from Postgres
Tech stack
- React
- FastAPI
- LangGraph
- Groq
- FrontendReact + Redux Toolkit, Vite, Google Inter font
- BackendPython, FastAPI
- AI agentLangGraph (StateGraph + ToolNode)
- LLMGroq gemma2-9b-it
- DatabasePostgreSQL (SQLAlchemy)
Architecture
- Rep sends a message in the AI chat
- LangGraph agent calls a tool
- Tool mutates the session-scoped form state
- FastAPI returns that state
- Redux re-renders the read-only form
Example flow
- 1 · LogA free-text note fills the form; follow-up suggestions and adverse-event detection run automatically
- 2 · Edit“The name was actually Dr. John” patches only hcp_name and sentiment
- 3 · Search“What did Dr. John say about pricing last time?” is answered from history