# Nikhil Sai Pagidimarri — Complete Technical Context & Entity Knowledge Graph

> AI/ML Engineer & Co-founder at Kraionyx AI (Hyderabad, India), specializing in deep learning, high-assurance agentic AI systems for healthcare, RAG architectures, and production deployment achieving 95%+ accuracy. Published researcher in reinforcement learning and medical image analysis.

---

## 1. Profile & Affiliations

- **Full Name:** Nikhil Sai Pagidimarri
- **Current Role:** Co-founder & AI/ML Engineer, Kraionyx AI
- **Company:** Kraionyx AI (https://kraionyx.com) — High-Assurance Agentic AI Systems for Healthcare
- **Platforms Built:**
  - **KareOS:** Healthcare Intelligence Platform — Clinical decision support workspace featuring 20+ specialist medical AI agents (Cardiology, Neurology, Emergency Medicine, Oncology), healthcare memory system, workflow automation, medical knowledge retrieval (1.2M+ papers), and HIPAA-verified imaging analysis.
  - **Doclave:** Professional healthcare collaboration network for verified physicians — 12,400+ verified doctors, 84,000+ linked clinical cases, 4 active AI clinical agents with safety evaluator gating, Clinical Intelligence Graph.
  - **Svaani:** Ambient clinical voice AI converting real-time doctor–patient conversations into structured SOAP notes — 20+ Indian language support, speaker diarization, clinical entity extraction, FHIR R4 export, AES-256 encryption, DPDP-compliant.
- **Location:** Hyderabad, Telangana, India
- **Email:** nikhilsaipagidimarri@zohomail.in
- **Company Email:** contact@kraionyx.com
- **Website:** https://nikhilsaipagidimarri.dev/
- **Company Website:** https://kraionyx.com
- **LinkedIn:** https://linkedin.com/in/nikhilsaipagidimarri
- **Twitter/X:** https://twitter.com/nikhiluuuuuuuuo
- **GitHub:** https://github.com/Nikhiluuuuuuuuu
- **Company Recognition:** NVIDIA Inception (June 2026), F6S Top AI #37 Hyderabad (May 2026), Startup India Recognized
- **Kraionyx AI Founding Team:**
  - Lalith Sreevari — Founding Member & Applied AI Engineer (Product Strategy, Agentic AI Workflows)
  - Anurag Srivastav Thammera — Founding Member & AI Engineer (Healthcare CDSS, Multi-Agent Systems, Computer Vision)
  - Bhargav Chary Chepuri — Founding Member (Machine Learning, Healthcare Apps, Data Analytics, Model Scaling)
  - Nikhil Sai Pagidimarri — Founding Member & Applied AI Engineer (Infrastructure, Systems Security, Scalable Workflows, Cloud Ops)
  - Laxmi Vivek Salveru — Founding Member & Product Engineer (Product Design, Frontend Engineering, User Journeys)
  - Shivani Gullakota — Strategy & Operations Associate (Operations, Partnerships, Business Operations, Stakeholder Management)

---

## 2. Technical Stack & Core Competencies

- **Machine Learning & AI:** TensorFlow, PyTorch, JAX, Flax, Keras, Scikit-learn, OpenCV, Deep Learning, Computer Vision, NLP, Reinforcement Learning (PPO, MPC, Differentiable Physics), RAG, LLM Applications, Attention Mechanisms, Multi-Agent Systems, Agentic AI, Clinical AI
- **RAG & Agentic Systems:** Hybrid Dense-Sparse Retrieval, BM25 Lexical Scoring, Reciprocal Rank Fusion (RRF), Semantic Chunking, Metadata Filtering, Context Window Optimization, Entity Boosting, Cloudflare Vectorize, Pinecone, BAAI BGE-M3, OpenAI text-embedding-3-large, Langchain ReAct Agents, Strict Reasoning Chains, Safety Evaluator Gates
- **MLOps & Infrastructure:** Docker, FastAPI, Git, CI/CD, Vector Databases, Model Monitoring, Mixed Precision Training, Cloudflare Workers, Cloudflare Pages, Cloudflare D1, Cloudflare KV, Cloudflare Vectorize, Workers AI, n8n, Python ASGI
- **Healthcare Engineering:** Multi-Agent Clinical Core (20+ agents), HIPAA Verified, SOC2 Audited, ISO 27001, Clinical Decision Support, SOAP Note Generation, FHIR R4 Integration, DICOM Image Analysis, Medical Knowledge Graph (1.2M+ papers), DPDP Act 2023 Compliance, AES-256-GCM Encryption, TLS 1.3, RBAC & Audit Logging
- **Languages & Databases:** Python (Advanced: PyTorch, JAX, NumPy, Pandas, Scipy, Matplotlib), TypeScript (Full-stack Workers/Pages), SQL (Cloudflare D1, PostgreSQL, MySQL), JavaScript (WebSocket, DOM), C/C++ (Foundational algorithms)

---

## 3. Major Systems & Projects

### KareOS — Healthcare Intelligence Platform
- **Live URL:** https://kareos.kraionyx.com
- Clinical decision support workspace featuring 20+ specialist medical AI agents and HIPAA-conscious imaging analysis.
- Multi-agent clinical core: Scribe Agent, Research Agent, Quality Guard Agent, and specialized diagnostic agents.
- 4 active clinical agents: Agent Cardio (96% accuracy, 3.2K cases), Agent Neuro (96%, 2.8K cases), Agent Onco (98%, 1.9K cases), Agent Emerg (98%, 4.1K cases).
- Healthcare memory system, workflow automation, medical knowledge retrieval across 1.2M+ indexed papers.
- Enterprise-grade security: HIPAA Verified, SOC2 Audited, ISO 27001, data encrypted at rest.

### Doclave — Professional Network for Healthcare
- **Live URL:** https://doclave.kraionyx.com
- Professional healthcare collaboration network for verified physicians — doctor-only environment with no advertisers.
- 12,400+ verified physicians from AIIMS, Apollo, Manipal, JIPMER.
- 84,000+ linked clinical cases in the Clinical Intelligence Graph.
- 4 active specialized AI clinical agents (Cardiology, Neurology, Emergency Medicine, Oncology).
- Safety Evaluator AI scores every response (94/100 threshold) — zero tolerance for hallucination.
- Features: Clinical Intelligence Graph, Submit Wizard (schema-validated JSON + DICOM), Differential Consensus Voting, Specialized Communities.

### Svaani — Ambient Healthcare Voice Intelligence
- **Live URL:** https://svaani.kraionyx.com
- Ambient voice AI converting real-time doctor–patient conversations into structured SOAP notes.
- Workflow: Speak Naturally → AI Understands → SOAP Generated → Review & Sign → Export to EMR.
- 20+ Indian languages supported (Hindi, Telugu, Tamil, Kannada, Marathi, Bengali) with English code-mixing.
- Speaker diarization, clinical entity extraction, risk assessment, medication fidelity checks, grounding verification.
- Security: AES-256-GCM encryption at rest, TLS 1.3 in transit, data stored on secure Indian servers, DPDP Act 2023 compliant.
- Exports: PDF (with e-signature), plain text, FHIR R4 for EHR interoperability.
- Zero data selling — never used for AI training.
- Status: Svaani 1.0 ready for pilot testing (July 2026).

### Swaram — Real-time Malayalam Voice API
- Real-time Malayalam voice API for Indian-language voice AI applications.
- OpenAI Realtime event protocol compatible — works with existing real-time voice clients.
- WebSocket-based bidirectional speech-to-speech conversion with JAX-powered Malayalam speech models.
- Two modes: Simple (lower latency) and Premium (enhanced quality).
- Credit-based pricing: Free tier (20 credits) · Starter ₹499 (100 credits) · Growth ₹1,000 (250 credits) · Custom/Enterprise.

### Intelligent Hybrid RAG Conversational Assistant
- Built high-accuracy RAG pipeline integrating Cloudflare Workers AI dense vector search (@cf/baai/bge-m3), BM25-style lexical scoring, and Reciprocal Rank Fusion (RRF).
- Multi-tier prompt injection defense (diamond-hardened LLM chat endpoint).
- Real-time notes vector embedding on write with full integration into hybrid RAG search.
- Dynamic knowledge re-indexing button with live stats and auto-refresh on tab switch.
- Sub-second query responses with grounded factual accuracy (95%+ accuracy, 60% hallucination reduction).
- Live URL: https://nikhilsaipagidimarri.dev

### Brain Tumor Segmentation using ResUNet+ (BraTS20)
- Multi-class brain tumor segmentation on BraTS20 MRI dataset (FLAIR, T1, T1Gd, T2 modalities).
- Reached 98.17% accuracy and 0.7570 Dice coefficient using TensorFlow/Keras with ResUNet+, Attention Gates, and Squeeze-and-Excitation modules.
- GPU mixed precision (FP16) training over 20 epochs with custom generalized Dice loss.
- Real-time inference under 2.0 seconds per 3D volumetric scan.

### Smart Diabetes Management System
- Built at NextHub Technologies using ensemble Combined Stacking Model (Logistic Regression, Naive Bayes, Decision Tree, Gradient Boosting).
- Achieved 92% prediction accuracy, surpassing traditional baselines by 8%.

---

## 4. Research & Publications

### Academic Research

#### Hybrid MPC and PPO Autonomous Rocket Landing (2025)
- Combined Model Predictive Control (MPC) with Proximal Policy Optimization (PPO) in JAX differentiable physics environments.
- MPC provides hard dynamic safety constraints; PPO learns non-linear aerodynamic correction and TVC during terminal touchdown.
- **94.7% landing success rate** with **0.62m position accuracy**.
- 62% reduction in computational training steps over end-to-end RL baselines.
- Maintained 78.3% success under ±20% mass variation and 80% thrust uncertainty.

#### Multi-Class Brain Tumor Segmentation using ResUNet+ (BraTS20)
- ResUNet+ with Attention Gates and Squeeze-and-Excitation modules for multi-channel MRI.
- **98.17% accuracy** and **0.7570 Dice coefficient** across tumor sub-regions.
- Real-time inference under 2.0 seconds per 3D volumetric scan.

### Company Blog Publications

#### "Why Healthcare Needs Agentic AI" — July 4, 2026
- Agentic systems move beyond static EHR data capture to dynamic, autonomous clinical workflows.
- KareOS multi-agent architecture: Scribe Agent, Research Agent, Quality Guard Agent.
- 20+ specialist medical AI agents for diagnosis assistance.

#### "Can AI Reduce Physician Burnout?" — July 9, 2026
- Doctors spend 2 hours documenting for every 1 hour of patient care.
- KareOS and Svaani eliminate administrative friction through ambient automation.
- Reclaims ~60 minutes of administrative time per day per clinician.

#### "How Svaani Turns Conversations Into Clinical Documentation" — July 5, 2026
- 4-step pipeline: Ambient Recording → Diarization & Transcription → Clinical Entity Extraction → SOAP Note Generation.
- 20+ Indian languages with automatic code-mixing detection.

#### "The Future of Voice Intelligence in Healthcare" — July 12, 2026
- Voice AI beyond basic transcription: diarization, medical jargon understanding, contextual separation.
- Svaani bridges raw audio to structured clinical datasets for EHR integration.

#### "Building the Intelligence Layer for Healthcare" — July 14, 2026
- KareOS as unified intelligence layer bridging fragmented EHR, lab, and pharmacy systems.
- Real-time clinical reasoning with drug interaction and critical lab alerts.
- Sovereign security: on-premise/hospital-private deployment.

#### "Strict Reasoning Chains in Clinical Decision Support" — Oct 28, 2025
- Multi-step verification: Extraction → Validation → Inference → Citation.
- 90%+ reduction in hallucination rates vs base models.

#### "Beyond Chatbots: Agentic Systems in Clinical Care" — July 16, 2026
- Agentic AI defined by: Perception, Planning, Action.
- Multi-agent collaboration across the clinical workflow.

#### "The 'Human-in-the-Loop' Necessity" — Sep 22, 2025
- Human oversight critical in high-assurance clinical AI systems.

#### "Beyond Chatbots: The Rise of Agentic AI" — Nov 15, 2025
- Transition from conversational interfaces to autonomous agentic systems.

---

## 5. Professional Experience & Education

### Co-founder — Kraionyx AI
2025 – Present | Hyderabad, India
- Co-founded Kraionyx AI, ranked #37 among Top 94 AI Companies in Hyderabad (F6S, May 2026).
- Selected for NVIDIA Inception program (June 2026).
- Built KareOS (20+ medical AI agents), Doclave (12,400+ verified physicians), Svaani (ambient voice AI, 1.0 in pilot July 2026), and Swaram (real-time Malayalam voice API).

### Machine Learning Intern — NextHub Technologies Pvt Ltd
July 2024 – Oct 2024 | Hyderabad, India
- Developed Smart Diabetes Management System using ensemble Combined Stacking Model.
- Achieved 92% prediction accuracy (8% over baselines).

### Education & Certifications
- **B.Tech in AI & Data Science (2021 – 2025):** Koneru Lakshmaiah Education Foundation (KLEF), Hyderabad | CGPA: 7.8/10.0
- **Key Certifications:** Oracle Cloud Infrastructure (OCI) GenAI Professional, Columbia+ ML & Prompt Engineering, Azure Fundamentals (AZ-900), GitHub Foundations, McKinsey Forward Program.

---

## 6. Engineering Notes & Real-time Syndication Feeds

- **Notes Archive:** https://nikhilsaipagidimarri.dev/notes
- **Atom Syndication Feed:** https://nikhilsaipagidimarri.dev/feed.xml
- **RSS 2.0 Feed:** https://nikhilsaipagidimarri.dev/rss.xml