👋 Hi there
🧑💻 Hi, I'm Rumpa (codekunoichi)
🌸 CTO building intelligent systems for ambulatory healthcare
📚 Deeply studying medical billing, coding, and CMS reimbursement policy to understand the system end-to-end
🧠 Exploring knowledge graphs on my DGX-Sparc workstation (NPI networks, payer patterns, provider clusters)
⚙️ Applying AI + policy + ERA data to shift denial prevention left
📈 100+ day GitHub streak — I learn by building every single day
Rolling 60-day contribution trend — consistency over time.
Here are some projects I’m actively working on 👇
- dgx-spark-open-source – DGX Spark setup notes, Ubuntu + GPU workflows, and cheatsheets.
- clinical_insights_ai – Clinical data exploration & AI experiments.
- demo-over-deck – Using AI-assisted MVPs to accelerate venture learning (demo-first mindset).
- fastai-hf-deployment – End-to-end playbook: train fastai locally → deploy to Hugging Face Spaces.
- fastai-medical-ml – fast.ai-inspired ML exercises applied to real ambulatory healthcare problems.
While some of my work lives in private repositories, here are the core systems I’m actively building today under the ContextRCM umbrella — focused on shift-left denial prevention, ERA-driven intelligence, and AI-assisted revenue cycle automation.
Although the initial pilots are in Behavioral Health, the architecture is fully specialty-agnostic and can extend to primary care, multi-specialty, surgical groups, and other ambulatory settings.
-
npi-intelligence-graph
End-to-end NPI analytics and knowledge-graph engine running on my DGX-Sparc workstation.
Transforms the national NPI registry into market intelligence and targeted lead-generation across any specialty. -
context-rcm-rules-engine
Unified rules engine combining CMS policy (NCCI, MUE, LCD/MCD), payer-specific quirks, and dynamic ERA-based behavior patterns to power shift-left denial prevention across specialties. -
context-rcm-denial-intelligence
ERA-driven denial clustering, preventability scoring, and “What-if” pre-submission simulation.
Core engine for proactive denial prevention in ambulatory workflows — not limited to BH. -
context-rcm-underpayment-detector
Statistical reconstruction of payer fee schedules using ERA payments.
Detects silent underpayments, shortfalls, and contract leakage for any CPT/modifier set.
📝 These repositories are private, but together they form the backbone of the ContextRCM mission: a proactive, intelligence-driven approach to ambulatory revenue cycle management — reducing denials, reconstructing fee schedules, enabling clean claims on first submission, and scaling across specialties with transparent, data-driven insight.
I’ve recently fallen down the rabbit hole of graph databases — and it turns out they’re a surprisingly natural fit for healthcare. I’ve been exploring how relationships actually behave in the real world: providers, payers, rules, denials, claims, and all the hidden links in between.
This curiosity led me to start sketching an RCM Knowledge Graph that connects:
- Payers
- CMS/NCCI/MUE/LCD rules
- Denial patterns (CARC/RARC)
- Claim histories
- Preventability clusters
- Underpayment signals with statistical significance
The idea is simple: let the graph surface insights that spreadsheets quietly bury.
I’ve been following along with Neo4j’s GraphAcademy courses as I learn how to design knowledge
graphs from first principles:
👉 Neo4J GraphAcademy Profile
I'm having so much fun connecting dots - literally - and it feels like a whole new way to think about revenue cycle intelligence.
- Languages: Python, JavaScript
- Cloud: AWS (RDS, S3, Lambda, DMS, HealthLake)
- AI/ML: Claude Code, Agentic AI, Fast.ai, scikit-learn
- Healthcare IT: EHR, RCM, HL7, FHIR
- Security & Compliance: HIPAA-ready architecture, PHI data isolation, RBAC design
- Sustaining a daily coding streak → learning by building every day
- Studying medical billing, coding, and CMS reimbursement rules
- Developing ContextRCM’s shift-left denial prevention architecture
- Building NPI-driven knowledge graphs on DGX-Sparc
- Exploring ERA intelligence for clean-claim automation