Varun Sai Raigir

I'm an AI/ML engineer focused on systems that reach real users, not just benchmarks. My research in clinical machine learning is published, and day to day I'm building edge AI for health wearables and a clinical documentation platform physicians already rely on. That experience is clinical, but my interest isn't limited to it: I'm equally comfortable across general software and AI engineering, from backend systems and APIs to production ML pipelines, and I'm open to roles beyond healthcare. What draws me in is the harder version of the problem, whether that's ML meeting constrained hardware, a regulated environment where getting it wrong has real consequences, or just a system that needs to actually ship and hold up in production. Originally from Hyderabad, India, now based in New Jersey.

1. AI/ML Engineer · INVIZA Health

   Jan 2025 to Present

   Building production edge AI algorithms for smart insole wearables: step count detection, gait analysis, and fall detection using sensor fusion and lightweight ML models compiled with TFLite Micro for on-device, real-time inference without cloud dependency. Own the full pipeline from raw IMU sensor data collection through model training in PyTorch and TensorFlow, quantization-aware optimization, and embedded deployment on constrained hardware. Benchmark accuracy and latency trade-offs across quantization levels to hit target performance budgets.

   • PyTorch
   • TensorFlow
   • TFLite Micro
   • IMU sensor fusion
   • Embedded C/C++

2. AI/ML Engineer · mednabdh.ai

   Mar 2025 to Aug 2026

   Designed and deployed a clinical AI platform now in active pilot with physicians. Real-time speech-to-text transcription, LangGraph-orchestrated multi-step LLM reasoning, RAG retrieval over a Pinecone vector store, structured SOAP note generation, and automated EMR field population via REST API. Physicians accept output with zero manual edits. Built the eval infrastructure that made clinical deployment possible: offline golden-set evaluation against clinician-reviewed cases, LLM-as-judge scoring against clinical documentation standards, CI-integrated regression tests gating every pipeline version, and responsible AI guardrails catching hallucinations before output reaches physicians. Deployed on Azure and GCP using Vertex AI and Gemini APIs, with Docker, CI/CD, structured logging, observability dashboards, role-based access controls, and audit logging for PHI compliance.

   • Python
   • FastAPI
   • PostgreSQL
   • React
   • LangChain
   • LangGraph
   • MCP servers
   • Pinecone
   • Vertex AI
   • Gemini
   • Azure
   • GCP
   • Docker

3. ML Research Associate · University of South Florida

   Dec 2023 to Dec 2025

   Designed a Graph Attention Network ensemble for precision medicine treatment stratification on multi-omics cancer data, achieving 92% accuracy across patient subgroups by modeling genomic, transcriptomic, and clinical features as a graph rather than independent variables. Used PySpark for distributed data processing across multi-node GPU clusters. Built automated data extraction and preprocessing pipelines with Python, Selenium, and NoSQL to integrate heterogeneous clinical datasets under strict quality and provenance controls. Also taught Operating Systems, guiding students through implementing core OS commands in C and C++ on Linux, covering file systems, process scheduling, and shell scripting.

   • PyTorch Geometric
   • PySpark
   • Python
   • Selenium
   • NoSQL

4. IT Enterprise Services Intern · Regeneron Pharmaceuticals

   May 2024 to Aug 2024

   Automated ECR-to-Nexus container migration using Jenkins pipelines and Python scripts, cutting deployment-related costs by over 15% by removing manual steps from the release process. Built interactive Splunk dashboards monitoring AWS asset utilization and compliance metrics, giving infrastructure teams real-time visibility into cloud security posture. Used Terraform for infrastructure-as-code provisioning.

   • Jenkins
   • Python
   • Splunk
   • AWS
   • Terraform

5. Software Developer · Amadeus

   Feb 2022 to Aug 2023

   Built modular Java Spring Boot microservices with MySQL and MSSQL backends for enterprise travel clients, exposing REST APIs consumed by Angular frontends. Designed scalable API contracts and optimized query performance. Diagnosed and resolved core-level production issues, including memory leaks, thread deadlocks, and connection pool exhaustion, using heap dumps, thread dumps, and query execution plans. Led testing with JUnit and PyTest. Reduced production bugs by 10% and response time by 5%.

   • Java
   • Spring Boot
   • MySQL
   • MSSQL
   • Angular
   • JUnit
   • PyTest

6. Data Scientist Intern · Facedapter

   Jan 2021 to Jul 2021

   Developed a CNN-based deep learning system to detect mask-based spoofing and presentation attacks in biometric authentication, increasing model accuracy to 89% through iterative architecture tuning. Engineered and curated training datasets by aggregating diverse image sources, improving generalization to real-world attack scenarios.

   • CNN
   • PyTorch
   • Computer Vision

Other work

• E. Zhang, V.S. Raigir et al., "MADSurv: An Uncertainty-Aware Framework for Multimodal Cancer Survival Analysis."

  ACM International Conference on Bioinformatics, Computational Biology, and Health Informatics (ACM BCB), 2025

• V.S. Raigir, "Integrative Multi-Omics and Clinical Data Analysis for Predicting Recurrence and Survival in Uterine Cancer."

  M.S. Thesis, University of South Florida, 2025

Open to conversations about software engineering, AI/ML, or anything in between. I’m open to relocating anywhere in the US, or working remote. The fastest way to reach me is email.