TWIN HEALTH | Whole Body Digital Twin™ AI Platform

Architected by Sai Likhith Kanuparthi for Gianluca Rossi (Senior Director of AI)

GitHub
Targeting Senior AI Engineer (Digital Twin Platform) • Gianluca Rossi
Location: Houston, TX | Remote (Open to Relocation)
Cleveland Clinic 2-Year RCT Breakthrough (NEJM Catalyst, 2025)

Reversing Chronic Metabolic Disease with Living Cyber-Physical AI

Unlike static single-point ML models or lifelong $12,000/yr GLP-1 medications (Ozempic/Wegovy), the Whole Body Digital Twin™ simulates multi-organ metabolic responses in real-time, repairing insulin sensitivity, clearing hepatic steatosis, and restoring endogenous beta-cell function.

71.0%
12-Mo Remission (Twin)
vs 2.4% Usual Care
72.8%
24-Mo Durability
NEJM Catalyst RCT
4.0M
Events / Min Streaming
Kafka / Flink Partitioning
Part 11
FDA Compliance
Immutable HMAC Audit
Sai Likhith Kanuparthi
Sai Likhith Kanuparthi • 7+ YOE
M.S. in Computer Science (NYU Tandon)
GPA 3.69/4.0 • Indian Patent 202541026299
+1 (860) 620-4718 • sailikhithcse@gmail.com

The 4-Pillar Candidate Unfair Advantage for Twin Health

FDA Healthtech + GenAI + Streaming + Autoencoders
1. Eli LillyFDA 21 CFR Part 11
Dose Management Platform

Built clinical radiopharmacy platform for radioactive F-18 PET radiotracers in Alzheimer's trials. Dual-layer cryptographic audit logging (PG triggers + Loki) with 99.9% uptime.

2. AirbnbGenAI Platform
FacadeDriver Model Gateway

Unified 30+ LLMs with in-flight Microsoft Presidio PII redaction (12 entities), 40-field dynamic batching (16x throughput: 600 to 10k rows/run), and $180k/yr semantic caching.

3. Southwest4M Events/Min
Kafka Real-Time Streaming

Architected high-throughput Kafka streaming pipelines with sub-50ms latency, partitioning by entity ID to guarantee strictly ordered time-series delivery with zero loss.

4. Shell PLCPatent 202541026299
Deep Learning Autoencoders

Engineered time-series reconstruction autoencoders for multi-sensor anomaly detection. Authored Indian Patent on Modular Deep Learning & Cross-Domain Transfer Learning.

Interactive Clinical AI Agent Studio

Whole Body Digital Twin™ Clinical Agent & Graph-RAG Workbench

Experience the real production AI pipeline: In-flight Presidio PHI scrub → Multimodal Biometric Fusion → Hyperbolic Biomedical Graph-RAG → Typewriter LLM Agent CoT → FDA 21 CFR Part 11 SHA-256 Seal.

Agent StatusIDLE (READY)
Model EngineTwin-Llama-70B-Med
Execution Speed:
Autonomous Multi-Stage Pipeline ProgressReady to Execute
11. Presidio PHI De-ID
12 Entities Masked
22. Multimodal Fusion
CGM ODE + Biometrics
33. Poincaré Graph-RAG
Multi-Hop Traversal
44. Clinical LLM CoT
Twin-Llama-70B Med
55. Part 11 Audit Seal
SHA-256 Electronic Sig
Continuous Glucose Monitor (CGM) Telemetry
5-Min Sampling
Normal Zone: 70–140 mg/dL
12:0012:3013:0013:3014:0014:30
Glucose Velocity (dv/dt)
+1.84 mg/dL/min (Accelerating)
Heart Rate Variability (HRV)
42 ms (Sympathetic Tone Active)
Sleep & Autonomic Tone
89% (Optimal Baseline)
Active Insulin on Board
0.8 units equivalent (Endogenous)
Agent Chain-of-Thought (Reasoning Traces)
Deterministic RAG
1. Presidio PHI scrub verified: Redacted 3 direct identifiers (<NAME_1>, <DATE_TIME_1>, <MEDICAL_RECORD_1>). Zero data egress outside HIPAA boundary.
2. Telemetry Ingestion: CGM slope delta = +1.84 mg/dL/min over 30m window. Threshold exceeded (+1.5 mg/dL/min).
3. Hyperbolic Graph Traversal: Queried Neo4j Knowledge Graph across FoodOn:0021 -> SkeletalMuscle:GLUT4.
4. Deterministic Clinical Logic: Patient is on Metformin (AMPK activator). Additional pharmacological intervention contraindicated.
5. Non-Pharmacological Directive: A 15-20 min postprandial walk directly mobilizes GLUT4 transporters to the cell surface via mechanical muscle contraction, attenuating peak excursion by ~35-50 mg/dL without risk of hypoglycemia.
Whole Body Digital Twin™ Coach Output
Click "Run Clinical Agent Loop" above to watch real-time token streaming and multi-step inference...
Eval Gate: PASS (0.00 Hallucination)Cleveland Clinic RCT Grounded
Interactive 3-Tier Distributed Architecture

Whole Body Digital Twin™ System Architecture (HLD)

Click any subsystem node in the interactive architecture diagram below to inspect its dataflow contracts, latency SLAs, architectural tradeoffs, and enterprise provenance.

SCALE: 100K+ PATIENTS4.0M EVENTS/MIN
Tier 1: Telemetry & Ingestion
<50ms SLA
Tier 2: Knowledge Graph & RAG
Zero-Hallucination
Subgraph Grounding Pipeline:
Patient → FoodOn (Nutrient) → RxNorm (Drug) → LOINC (Biomarker)
Tier 3: FacadeDriver & Governance
FDA Part 11
Tier 2: Knowledge Graph & RAG

Poincaré Hyperbolic Graph-RAG Engine

SLA: <35ms multi-hop traversalThroughput: 10K graph traversals/sec
Architectural Role & Design Decisions:

Projects hierarchical biomedical trees into hyperbolic space (Poincaré ball disc), enabling deterministic multi-hop traversal from nutrients through drug pharmacokinetics to glycemic outcomes with zero hallucinations.

High-Consequence Engineering Tradeoff:

Tradeoff: Hyperbolic distance calculations require non-Euclidean Riemannian metric operations, optimized via vectorized C++ bindings.

Production Tech Stack:
Neo4j EnterprisePoincaré Hyperbolic EmbeddingsCypherNetworkX
Candidate Track-Record Provenance:Non-Euclidean Biomedical Graph Research
Subsystem A: Real-Time Telemetry Substrate

Continuous 5-minute CGM streams partitioned strictly by patient_id in Kafka to guarantee order. Apache Flink computes 15-minute sliding rate-of-change and feeds deep time-series autoencoders to detect acute metabolic anomalies.

Latency: <50msSouthwest & Shell
Subsystem B: Biomedical Knowledge Graph

Standardizes labs (LOINC), drugs (RxNorm), clinical phenotypes (SNOMED-CT), and nutrition (FoodOn). Uses Poincaré hyperbolic embeddings to preserve deep parent-child taxonomy hierarchies and multi-hop graph traversals to ensure zero hallucination.

Accuracy: DeterministicNeo4j Graph-RAG
Subsystem C: FacadeDriver & 21 CFR Part 11

FacadeDriver decouples 30+ LLMs with in-flight Presidio PII redaction and $180k/yr semantic caching. FDA 21 CFR Part 11 dual-layer audit trail (PostgreSQL triggers + Loki events) guarantees immutable, tamper-evident cryptographic compliance.

Governance: FDA Part 11Eli Lilly & Airbnb
Low-Level Component Design (LLD) & Execution Flows

Component Contracts, Data Schemas & Sequence Logic

Production-grade class specifications for asynchronous model orchestration, in-memory sanitization, and cryptographic audit trails.

Sequence Flow: End-to-End Sensor-to-Clinical Coaching Datapath
Sub-500ms End-to-End P99
STEP 01
CGM Telemetry Ingest

Patient CGM emits 5-min glucose (185 mg/dL). Kafka partitions strictly by patient_id.

STEP 02
Flink Velocity & AE

Flink calculates 15-min velocity (+0.4) and autoencoder checks reconstruction error ($L > \tau$).

STEP 03
Graph-RAG Traversal

Traverses causal graph: FoodOn:0021RxNorm:6809LOINC:4548.

STEP 04
FacadeDriver Runtime

Presidio in-flight PII redaction, Redis semantic cache check, and model dispatch with safety guardrails.

STEP 05
Part 11 Audit & Push

SHA-256 HMAC cryptographic audit logged. Real-time push delivered to patient mobile coaching app.

facade_driver_gateway.py
class FacadeDriverGateway:
    def __init__(self, redis_cache: RedisClient, presidio_engine: AnonymizerEngine):
        self.cache = redis_cache
        self.anonymizer = presidio_engine
        self.model_pool = AsyncWorkerPool(max_workers=64)
        
    async def execute_clinical_prompt(self, patient_id: str, graph_context: GraphContext, prompt: str) -> ClinicalResponse:
        # 1. In-Flight Presidio PII Redaction
        sanitized_prompt = self.anonymizer.anonymize(prompt, entities=HEALTHCARE_ENTITIES_12)
        
        # 2. Redis Semantic Cache Check
        cache_key = self.generate_embedding_hash(sanitized_prompt, graph_context.rules_hash)
        if cached := await self.cache.get(cache_key):
            return ClinicalResponse.from_cache(cached)
            
        # 3. Dynamic Model Routing & Execution
        response = await self.model_pool.dispatch_async(sanitized_prompt, context=graph_context)
        
        # 4. Deterministic Schema & Boundary Validation
        assert ClinicalGuardrails.validate_safety_boundaries(response), "Clinical boundary violation"
        await self.cache.set(cache_key, response, ttl=86400)
        return response
part11_audit_logger.py
class Part11AuditLogger:
    def __init__(self, db_session: AsyncSession, hmac_key: bytes):
        self.db = db_session
        self.hmac_key = hmac_key
        
    async def log_clinical_event(self, patient_id: str, event_type: str, old_val: dict, new_val: dict, actor: str):
        timestamp = datetime.now(timezone.utc)
        payload = json.dumps({"patient_id": patient_id, "old": old_val, "new": new_val, "ts": timestamp.isoformat()})
        
        # Cryptographic SHA-256 HMAC Signature for Tamper Evidence
        signature = hmac.new(self.hmac_key, payload.encode('utf-8'), hashlib.sha256).hexdigest()
        
        audit_record = AuditLog(
            patient_id=patient_id,
            action=event_type,
            old_value=old_val,
            new_value=new_val,
            performed_by=actor,
            performed_at=timestamp,
            hmac_signature=signature
        )
        self.db.add(audit_record)
        await self.db.commit()
First-Principles Engineering & Mathematical Algorithms

Core Data Structures & Algorithms (DSA Deep Dive)

First-principles algorithms for real-time sensor streams, non-Euclidean taxonomy manifolds, and zero-hallucination causal reasoning.

1. Sliding-Window Real-Time Glucose VelocityO(W) Time | O(1) Space

Computes the first derivative rate-of-change (dGdt\frac{dG}{dt}) over a streaming window WW using online Ordinary Least Squares (OLS) linear regression to catch glycemic spikes before they crest.

dGdt=Wi=1W(tiGi)(i=1Wti)(i=1WGi)Wi=1Wti2(i=1Wti)2\frac{dG}{dt} = \frac{W \sum_{i=1}^W (t_i \cdot G_i) - \left(\sum_{i=1}^W t_i\right)\left(\sum_{i=1}^W G_i\right)}{W \sum_{i=1}^W t_i^2 - \left(\sum_{i=1}^W t_i\right)^2}
def calculate_glucose_velocity(readings: list[tuple[float, float]]) -> float:
    # readings = [(t0, g0), (t1, g1), ... (tN, gN)]
    n = len(readings)
    sum_t = sum(t for t, g in readings)
    sum_g = sum(g for t, g in readings)
    sum_tg = sum(t * g for t, g in readings)
    sum_t2 = sum(t * t for t, g in readings)
    return (n * sum_tg - sum_t * sum_g) / (n * sum_t2 - sum_t ** 2)
2. Multi-Hop Causal Graph-RAG TraversalO(V + E log V) Dijkstra

Traverses directed acyclic metabolic paths from ingested nutrients (FoodOn) through medication pharmacokinetics (RxNorm) to biomarker targets (LOINC) with zero contraindications.

Score(P)=ePw(e)I(Contraindication(e)=)\text{Score}(P) = \prod_{e \in P} w(e) \cdot \mathbb{I}\Big(\text{Contraindication}(e) = \emptyset\Big)
def find_clinical_causal_path(graph: nx.DiGraph, food_node: str, biomarker_target: str) -> list[str]:
    # Priority queue based traversal enforcing zero contraindication constraints
    dist = {food_node: 0}
    pq = [(0, food_node, [food_node])]
    while pq:
        cost, curr, path = heapq.heappop(pq)
        if curr == biomarker_target:
            return path
        for neighbor in graph.neighbors(curr):
            if not graph[curr][neighbor].get('contraindicated'):
                new_cost = cost + graph[curr][neighbor]['weight']
                if new_cost < dist.get(neighbor, float('inf')):
                    dist[neighbor] = new_cost
                    heapq.heappush(pq, (new_cost, neighbor, path + [neighbor]))
    return []
3. Poincaré Hyperbolic Taxonomy EmbeddingsNon-Euclidean Manifold

Embeds deep biomedical tree hierarchies (SNOMED-CT & FoodOn) into Poincaré ball Dd\mathbb{D}^d where distance grows exponentially toward the boundary, preserving parent-child transitive depth without distortion.

dD(u,v)=arcosh(1+2uv2(1u2)(1v2))d_{\mathbb{D}}(u, v) = \operatorname{arcosh}\left(1 + 2 \cdot \frac{\|u - v\|^2}{(1 - \|u\|^2)(1 - \|v\|^2)}\right)
Why Hyperbolic Geometry: Euclidean embeddings fail when modeling tree-like ontology graphs because the circumference of an n-ary tree grows exponentially with depth (O(bd)O(b^d)), while Euclidean volume only grows polynomially (O(rd)O(r^d)).
4. Autoencoder Reconstruction Anomaly DetectionDeep Learning (Shell Patent)

Deep autoencoder trained on stable metabolic states projects continuous multi-sensor telemetry into latent bottleneck zz. Spikes in reconstruction error trigger automated clinical intervention alarms.

LMSE(x,x^)=1Di=1D(xix^i)2>τthreshold\mathcal{L}_{\text{MSE}}(x, \hat{x}) = \frac{1}{D} \sum_{i=1}^D (x_i - \hat{x}_i)^2 > \tau_{\text{threshold}}
Proven Track Record: Leveraged at Shell for real-time sensor anomaly detection across continuous sensor streams and documented in Indian Patent 202541026299.
Peer-Reviewed Clinical Trials & Economics

Clinical Evidence: Cleveland Clinic RCT & GLP-1 Disruption

Published in NEJM Catalyst (2025): Proving curative metabolic disease reversal vs lifelong pharmaceutical dependency.

Cleveland Clinic 2-Year RCT (NCT05181449)
71.0% Remission at 12 Months:

Patients achieved HbA1c < 6.5% with complete elimination of all diabetes medications (including insulin and sulfonylureas) vs only 2.4% in usual care.

72.8% Sustained at 24 Months:

Proves long-term durability of endogenous beta-cell healing, reversing the root pathology rather than masking symptoms.

84% Hepatic Steatosis Reduction:

Direct MRI-PDFF measurements demonstrated massive reductions in liver fat, resolving non-alcoholic fatty liver disease (NAFLD).

Twin Health vs Lifelong GLP-1 Medications (Newsweek Analysis)
$12,000 to $15,000 / yr
GLP-1 Cost (Lifelong)
Curative Remission
Twin Health Model
  • Lean Muscle Preservation: GLP-1s cause up to 40% weight loss from lean muscle mass. Twin Health restores metabolic rate and protects skeletal muscle.
  • 70% 2-Year Discontinuation: Severe GI intolerance causes high drop-out rates on GLP-1s with rapid rebound weight spikes.
  • Employer ROI: Twin Health pays for itself within 6 months by eliminating drug claims for self-insured employers.
7+ Years Track Record

Professional Experience & Engineering Provenance

Proven history building FDA-regulated clinical healthtech, enterprise GenAI inference engines, and 4M event/min real-time streaming platforms.

Airbnb

Senior Software Engineer, ML Infrastructure & AI Engineering (GenAI Platform)
Sep 2024 - Present
San Francisco, CA (Remote)
  • Architected FacadeDriver, unifying 30+ LLMs behind an asynchronous worker pool with in-flight Microsoft Presidio PII redaction across 12 healthcare entities, achieving 30% pipeline speedup and zero data leakage.
  • Scaled dynamic batching from 600 to 10,000 rows/run (16x throughput scaling) across 40 distinct input fields while sustaining 99.9% availability.
  • Engineered multi-tier Redis semantic caching layer, eliminating redundant inferences and saving $180k/year in compute costs while reducing P99 latency by 38%.
  • Constructed automated 23-version evaluation harness with 1,690 ground-truth benchmark cases, integrating LLM-as-a-Judge scoring release gates.

Eli Lilly and Company

Senior Software Engineer - Dose Management Platform (21 CFR Part 11)
Feb 2024 - Aug 2024
Philadelphia, PA / Indianapolis, IN (Remote)
  • Architected clinical radiopharmacy Dose Management System under FDA 21 CFR Part 11 regulations for radioactive F-18 PET radiotracers (Amyvid / Tauvid) used in Alzheimer's disease clinical trials (Donanemab context).
  • Engineered dual-layer audit trail combining database triggers on PostgreSQL and structured Loki business events, providing tamper-evident SHA-256 HMAC cryptographic verification.
  • Maintained 99.9% uptime for time-critical radiopharmaceutical distribution where radioactive decay (110-min half-life) demanded deterministic sub-minute execution.
  • Implemented real-time compliance alerting and automated rollback mechanisms for GxP validation across enterprise clinical sites.

Southwest Airlines

Senior Software Engineer - Backend & Data Platform
Jan 2023 - Jan 2024
Dallas, TX
  • Architected distributed real-time telemetry streaming pipelines on Apache Kafka, processing 4.0M events/min with sub-50ms latency.
  • Implemented strictly ordered partition routing keyed by entity ID, preventing out-of-order state corruptions across high-velocity time-series streams.
  • Reduced consumer lag by 65% by tuning Flink tumbling and sliding compute windows and parallelizing stateful stream processors.

Shell PLC

Senior Software Engineer - Backend & Data Science
Jun 2021 - Dec 2022
Houston, TX
  • Engineered deep learning autoencoder neural networks for continuous multi-sensor telemetry anomaly detection, identifying state drift prior to operational failure.
  • Authored Indian Patent (Application No. 202541026299) on Modular Deep Learning Architecture for Cross-Domain Transfer and Incremental Learning.
  • Built distributed model inference pipeline deployed across cloud endpoints with automated data drift monitoring and active retraining hooks.

Oracle

Software Engineer - ERP Analytics & Data Engineering
Aug 2017 - Jul 2019
Bengaluru, India
  • Engineered high-throughput ETL data pipelines ingesting transactional records across enterprise relational databases.
  • Optimized complex SQL query plans and partitioned table indexing, accelerating batch aggregation runtimes by 45%.
Open-Source Systems & Technical Honors

Open-Source Contributions, Patents & Education

Core architectural contributions to premier AI frameworks, deep learning patents, and academic credentials.

Open-Source Systems Contributions
LiteLLM (Google Vertex AI Proxy Fix):

Resolved critical authentication token header propagation across Google Vertex AI endpoints, preventing silent authorization drops on enterprise multi-model gateways.

LangChain (Agentic Search Cost Attribution):

Fixed search cost tracking across LangSmith and Braintrust by propagating query count metadata into agent execution contexts, eliminating 10x cost underreporting.

LiveKit Agents (Multi-Turn Voice Reliability):

Engineered multi-turn reliability telemetry hooks in voice agent evaluation pipelines and authored 23 unit test suites for observer state machines.

Patents & Technical Honors
Indian Patent Published:

Title: Modular Deep Learning Architecture for Cross-Domain Transfer and Incremental Learning

App No: 202541026299 (Indian Patent Office)
Industry Certifications:
  • Google Cloud Certified Professional Data Engineer
  • AWS Certified Solutions Architect • Professional (SAP-C02)
  • AWS Certified Machine Learning • Specialty (MLS-C01)
  • Microsoft Certified: Azure Data Scientist Associate (DP-100)
  • Google Foobar Challenge • Completed Level 3
Education & Academic Credentials
New York University (NYU Tandon)
M.S. in Computer Science
GPA: 3.69 / 4.0 • Brooklyn, NY • Sep 2019 - May 2021

Coursework: Distributed Systems, Deep Learning, Machine Learning, Cloud Computing, Algorithms.

Jawaharlal Nehru Technological University (JNTU)
B.Tech in Computer Science and Engineering
Hyderabad, India • Aug 2013 - May 2017
Immediate Engineering Velocity

30-60-90 Day Execution Blueprint for Twin Health

A proactive, candidate-owned technical execution roadmap designed to accelerate platform scalability and regulatory readiness.

Days 1 - 30

Telemetry Ingestion & Presidio Hardening

  • Audit existing CGM ingestion telemetry pipelines and benchmark Kafka partition lag across concurrent patient streams.
  • Implement in-flight Presidio PII/PHI redaction layer across all model gateways, guaranteeing 100% HIPAA/Part 11 compliance.
  • Establish baseline latency and cost telemetry dashboards across foundation model inference endpoints.
Days 31 - 60

Biomedical Graph-RAG & Poincaré Embeddings

  • Unify LOINC, RxNorm, SNOMED-CT, and FoodOn ontologies into a unified Neo4j causal metabolic knowledge graph.
  • Implement Poincaré hyperbolic tree embeddings to capture deep parent-child taxonomy transitive depth with zero distortion.
  • Deploy deterministic multi-hop causal constraint validation layer to eliminate hallucinations in patient coaching recommendations.
Days 61 - 90

FacadeDriver Multi-Model Runtime & Eval Gates

  • Deploy FacadeDriver asynchronous worker pool with 40-field dynamic batching and Redis semantic caching.
  • Construct automated 23-version evaluation harness with 1,690 ground-truth clinical trial cases for LLM-as-a-Judge gating.
  • Integrate full cryptographic SHA-256 HMAC audit logging for FDA 21 CFR Part 11 commercial readiness.
Direct Candidate Contact

Let's Build the Future of Living Cyber-Physical AI

Prepared for Gianluca Rossi (Senior Director of AI) • Available for immediate technical deep-dive.