Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

I’m a staff-level AI/ML engineer and technical lead based in Bucharest, focused on building production-grade agentic AI systems and voice-first assistants. Over the last 8 years in applied AI and 12+ years in software engineering, I’ve led end-to-end delivery across architecture, implementation, evaluation, and cloud deployment—especially where reliability, observability, and maintainability matter. I’ve delivered systems that combine real-time speech, multi-agent orchestration, and retrieval-augmented generation, including GraphRAG-style knowledge platforms and model-serving optimizations on GPU infrastructure. I enjoy collaborating with teams through architecture reviews and code reviews, using clear acceptance criteria and measurable evaluation to reduce operational friction and improve outcomes in production.…I’m a staff-level AI/ML engineer and technical lead based in Bucharest, focused on building production-grade agentic AI systems and voice-first assistants. Over the last 8 years in applied AI and 12+ years in software engineering, I’ve led end-to-end delivery across architecture, implementation, evaluation, and cloud deployment—especially where reliability, observability, and maintainability matter. I’ve delivered systems that combine real-time speech, multi-agent orchestration, and retrieval-augmented generation, including GraphRAG-style knowledge platforms and model-serving optimizations on GPU infrastructure. I enjoy collaborating with teams through architecture reviews and code reviews, using clear acceptance criteria and measurable evaluation to reduce operational friction and improve outcomes in production.

Daniel Stan Senior

AI Engineer, Data Scientist, DevOps Developer, +4





I’m a staff-level AI/ML engineer and technical lead based in Bucharest, focused on building production-grade agentic AI systems and voice-first assistants. Over the last 8 years in applied AI and 12+ years in software engineering, I’ve led end-to-end delivery across architecture, implementation, evaluation, and cloud deployment—especially where reliability, observability, and maintainability matter. I’ve delivered systems that combine real-time speech, multi-agent orchestration, and retrieval-augmented generation, including GraphRAG-style knowledge platforms and model-serving optimizations on GPU infrastructure. I enjoy collaborating with teams through architecture reviews and code reviews, using clear acceptance criteria and measurable evaluation to reduce operational friction and improve outcomes in production.…I’m a staff-level AI/ML engineer and technical lead based in Bucharest, focused on building production-grade agentic AI systems and voice-first assistants. Over the last 8 years in applied AI and 12+ years in software engineering, I’ve led end-to-end delivery across architecture, implementation, evaluation, and cloud deployment—especially where reliability, observability, and maintainability matter. I’ve delivered systems that combine real-time speech, multi-agent orchestration, and retrieval-augmented generation, including GraphRAG-style knowledge platforms and model-serving optimizations on GPU infrastructure. I enjoy collaborating with teams through architecture reviews and code reviews, using clear acceptance criteria and measurable evaluation to reduce operational friction and improve outcomes in production.

Available to hire

I’m a staff-level AI/ML engineer and technical lead based in Bucharest, focused on building production-grade agentic AI systems and voice-first assistants. Over the last 8 years in applied AI and 12+ years in software engineering, I’ve led end-to-end delivery across architecture, implementation, evaluation, and cloud deployment—especially where reliability, observability, and maintainability matter.

I’ve delivered systems that combine real-time speech, multi-agent orchestration, and retrieval-augmented generation, including GraphRAG-style knowledge platforms and model-serving optimizations on GPU infrastructure. I enjoy collaborating with teams through architecture reviews and code reviews, using clear acceptance criteria and measurable evaluation to reduce operational friction and improve outcomes in production.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Intermediate

Intermediate

Intermediate

Intermediate

Intermediate

Beginner

Beginner

Beginner

Language

English

Advanced

Romanian, Moldavian, Moldovan

Fluent

Work Experience

Senior AI/ML Engineer at HCA Healthcare Tennessee

January 1, 2024 - September 1, 2026

Led architecture and hands-on delivery of a clinical, voice-first assistant and GraphRAG knowledge platform on AWS. Built and orchestrated LangGraph agents using Bedrock/AgentCore and MCP tools with explicit state management, bounded retries/timeouts, and recovery paths. Engineered a Neo4j GraphRAG pipeline with entity normalization, metadata filtering, and re-ranking, improving retrieval precision at top-k results. Integrated real-time streaming speech and synthesis (Deepgram STT, custom TTS, LiveKit) with endpointing and interruption handling to keep conversation latency low. Optimized GPU inference using vLLM/CUDA (PagedAttention, KV/prefix caching) to reduce p95 serving latency. Implemented validated Databricks processing aligned with FHIR and improved valid downstream payload rates through schema/terminology/required-field checks. Standardized Docker builds and GitHub CI/CD with versioned artifacts and rollback checks, reducing model release lead time. Improved operational excelle

Senior AI/ML Engineer | Technical Lead at Tandem Health

January 1, 2024 - Present

Led architecture and technical system design of a real-time voice-first healthcare AI system using AWS Bedrock and agent tooling (AgentCore, LangChain, LangGraph). Implemented FastAPI APIs in Docker on AWS to improve conversation speed and reliability. Built a graph-based RAG system with Neo4j for structured clinical retrieval and multi-hop reasoning. Designed and integrated MCP tools for standardized, secure access to external services, knowledge sources, patient data, and internal APIs for dynamic multi-step workflows. Fine-tuned domain-specific healthcare LLMs for reasoning accuracy, safety alignment, and dialogue consistency. Developed a real-time voice pipeline using LiveKit, Deepgram (STT), and custom TTS for sub-second responsiveness. Established AWS end-to-end MLOps with model versioning, CI/CD, and automated retraining; added Prometheus/Grafana observability; optimized inference with vLLM and CUDA techniques; and contributed to Next.js/TypeScript frontend integration.

AI Software Engineer at Amdocs Limited Missouri

January 1, 2020 - December 1, 2023

Delivered AI-enabled telecom operations and engineering applications spanning anomaly detection and document intelligence. Built a telecom anomaly-detection platform connecting ML services to production backends for analysis of 20+ network elements, reducing manual network data inspection time by 34%. Developed FastAPI/React/TypeScript applications with typed API contracts for anomaly investigation and engineering search, reducing median investigation time from 28 minutes to 18 minutes. Implemented Azure and Databricks ingestion pipelines with schema and data-quality checks, reducing dataset preparation time from 9 hours to 5 hours for recurring processing runs. Built PyTorch/ONNX vision pipelines for antenna and telecom-equipment analysis, improving detection precision by ~22%. Created FAISS and Pinecone semantic/vector-search RAG services integrated with backend APIs, reducing search time by over 60%. Deployed Docker/Kubernetes services with health probes, resource controls, scalable

AI Software Engineer at Amdocs Limited

January 1, 2020 - December 1, 2023

Developed AI-enabled telecom operations solutions spanning anomaly detection and document intelligence. Built a telecom anomaly-detection platform integrating ML services with production backends for analysis across 20+ network elements and reducing manual inspection time. Delivered FastAPI/React/TypeScript applications with typed API contracts to accelerate anomaly investigation and engineering search. Implemented Azure/Databricks ingestion pipelines with schema and data-quality checks to reduce dataset preparation time. Created PyTorch/ONNX vision pipelines for antenna and telecom equipment analysis (including 3D bounding boxes) and improved detection precision. Built FAISS/Pinecone semantic retrieval (RAG) services integrated into backend APIs and engineering tools, reducing search time substantially. Deployed Docker/Kubernetes services with telemetry and scalability controls, reducing peak-load failures in stress tests; contributed to architecture reviews and documentation, reducin

Artificial Intelligence Engineer at Amdocs Limited

January 1, 2020 - December 1, 2023

Developed a large-scale telecom anomaly detection system for T-Mobile to reduce manual inspection workload and operational costs. Built data pipelines in Databricks and Azure to transform unstructured engineering documents into structured datasets for downstream analytics. Created document-intelligence solutions using Azure Document Intelligence and Hugging Face LLMs to automate key information extraction and improve accuracy. Developed 3D bounding box detection models with PyTorch/ONNX (reported ~22% precision improvement). Built hybrid search using FAISS and Pinecone to reduce search time by 60%+. Implemented NLP pipelines (Python, spaCy, FAISS) for classification, entity extraction, and discrepancy detection to reduce manual review. Implemented scalable MLOps on Azure Machine Learning with automated testing and CI/CD to speed releases.

Machine Learning Engineer at Luxoft Bucharest

September 1, 2018 - December 1, 2019

Worked on applied ML analytics for automotive driver monitoring and ADAS vision. Developed CNN/RNN driver-monitoring models in TensorFlow and PyTorch using 500+ labeled video hours across 12 drivers for distraction detection and safety alerts. Optimized ADAS computer-vision pipelines, reducing GPU inference latency by 30% through profiling and preprocessing improvements. Built LSTM and Transformer forecasting models with lagged and multivariate features, reducing forecast MAE by 17% versus baseline on held-out time windows. Implemented K-means customer segmentation, improving targeting and increasing campaign conversion rates by 20%. Created Tableau dashboards for reporting on model quality, conversion lift, and pipeline latency, reducing weekly report preparation from 3 hours to 45 minutes. Documented validation results and integration requirements, reducing model-handover clarification cycles from 5 to 3 review rounds.

Machine Learning Engineer at Luxoft

September 1, 2018 - December 1, 2019

Developed a driver-monitoring system using CNN/RNN models in TensorFlow and PyTorch, integrating real-time video analysis to detect distraction and improve safety alerts. Built ADAS computer vision pipelines, reducing inference latency by 30%. Designed recommendation algorithms leveraging Elasticsearch for personalized content. Developed anomaly-detection models using FAISS and RNN architectures to reduce system failures by 10%. Deployed ML services on AWS using Docker and Kubernetes for reliable continuous updates.

Software Engineer at Zitec Bucharest

August 1, 2012 - July 1, 2016

Developed modular Python backend services using Django and Flask for multiple products, reducing median delivery time for comparable features from 8 to 6 working days. Optimized PostgreSQL/MySQL indexing and query plans for applications serving ~20,000 daily requests, reducing average latency for frequently used queries from 240ms to 145ms. Implemented Celery with RabbitMQ for background processing (notifications and data pipelines), increasing throughput from 420 to 690 tasks per minute. Standardized deployment scripts, environment configuration, and rollback checks, reducing routine release execution time from 75 minutes to 30 minutes. Investigated production errors and documented repeatable fixes, reducing median resolution time for recurring issues from 4 hours to 2.5 hours.

Software Engineer at Zitec

August 1, 2012 - July 1, 2016

Developed backend services using Python, Django, and Flask with modular components to support core product features, enabling faster rollout and improved reliability. Optimized PostgreSQL/MySQL performance through query refactoring and indexing to reduce average latency. Built RESTful APIs and integration services using Django REST Framework to streamline data exchange and reduce integration errors. Implemented asynchronous background processing with RabbitMQ and Celery for tasks such as email notifications and data pipelines, increasing throughput and freeing web request threads.