Submission type
Coded Agent with UiPath SDK
Name
Sai Kiran Kondreddy
Industry category in which use case would best fit in (Select up to 2 industries)
Banking
Finance
Healthcare Pharma
Insurance
Complexity level
Intermediate
Summary (abstract)
The UiPath LangChain Redaction Agent is a coded specialist agent designed to automatically detect and redact Protected Health Information (PHI) and Personally Identifiable Information (PII) from healthcare and enterprise datasets. It ensures HIPAA Safe Harbor compliance by using UiPath’s LLM and LangChain SDK to intelligently mask sensitive fields such as names, MRNs, phone numbers, emails, and dates of birth before sharing data with downstream automations.
This agent acts as a privacy firewall for automation pipelines, enabling organizations to run analytics, testing, and AI workloads on de-identified data — ensuring trust, compliance, and explainability
Detailed problem statement
Healthcare and insurance workflows often require accessing patient or member records across databases for analytics, reporting, and test data generation. These databases contain PHI/PII governed by HIPAA.
Current redaction approaches are limited:
Manual masking is inconsistent and slow.
Regex-based tools lack context and fail on unstructured data.
Legacy scripts are not explainable or traceable within automation pipelines.
Organizations needed a governed, reusable, and context-aware redaction agent that could integrate directly into UiPath workflows and scale across databases — without violating compliance.
Detailed solution
The UiPath LangChain Redaction Agent addresses this challenge through a modular, LangGraph-based coded agent architecture:
Natural Language → SQL Node
Converts user or process queries into optimized SQL using UiPath’s integrated LLM.
Fetch Data Node
Executes queries securely via SQLAlchemy; retrieves PHI/PII records from any database (SQLite, PostgreSQL, MySQL).
Redact Data Node
Applies UiPath LLM-driven HIPAA Safe Harbor masking rules:
Names → PATIENT_
DOB → DOB_
MRN → MRN_
Phones/Emails/Addresses masked contextually.
JSON Output Node
Returns clean, auditable data and redaction policy metadata to the caller.
Architecture: START → NL-to-SQL → Fetch → Redact → END
Deployment: Published as a Coded Agent in UiPath Automation Cloud; callable from Studio, Orchestrator, or REST API.
Observability: Built-in UiPath tracing, logging, and LangSmith event tracking.
Narrated video link (sample: https://bit.ly/4pvuNEL)
Expected impact of this automation
Compliance Guarantee: End-to-end HIPAA Safe Harbor enforcement.
Productivity Boost: Zero manual intervention; redaction automation cuts processing time by >80%.
Trust & Transparency: Every output includes redaction policy and decision rationale.
Scalability: Multi-database and multi-tenant ready with cloud observability.
Reuse & Extendibility: Serves as a reusable privacy micro-agent for any UiPath process needing PHI/PII masking.
UiPath products used (select up to 4 items)
UiPath Coded Agents
UiPath Orchestrator
Integration with external technologies
Databases: PostgreSQL, MySQL, SQLite via SQLAlchemy. Environment Config: .env via python-dotenv or UiPath Assets. LLM Stack: OpenAI GPT-4o models through UiPath LLM Interface. Data Pipeline Integration: Invocable from external analytics or test-data generation bots via REST or UiPath Apps. Compliance Validation: Aligned with HIPAA Safe Harbor and PHI/PII classification standards.
Hi there, 
Nov 5–16: Jury evaluation by 
Nov 17: Winners announced 
.
,