Robert Pavel - Senior AI/ML Engineer

Q: Where is Robert based?

Robert is based in Bucharest, Romania .

Q: What languages does Robert speak?

Robert speaks the following languages: English (Advanced), Romanian (Advanced) .

Q: How many years of experience does Robert have?

Robert has at least 10 years of experience. During this time, Robert has worked in at least 4 different roles and for 4 different companies . The average length of individual experience is 2 years and 5 months . Note that Robert may not have shared all experience and actually has more experience.

Q: What is Robert's latest experience?

Robert's most recent position is AI Platform Engineer at Wolters Kluwer .

Q: What companies has Robert worked for in recent years?

In recent years, Robert has worked for Wolters Kluwer and Sage .

Q: Which industries is Robert most experienced in?

Robert is most experienced in industries like Information Technology , Banking and Finance , and Retail . Robert also has some experience in Healthcare and Pharmaceutical .

Q: Which business areas is Robert most experienced in?

Robert is most experienced in business areas like Information Technology , Business Intelligence , and Research and Development . Robert also has some experience in Finance and Investments and M&A .

Q: Which industries has Robert worked in recently?

Robert has recently worked in industries like Banking and Finance , Information Technology , and Healthcare .

Q: Which business areas has Robert worked in recently?

Robert has recently worked in business areas like Information Technology , Business Intelligence , and Finance .

Recommended expert

Bucharest, Romania

Check rate

Experience

Jul 2024 - Apr 2026

1 year 10 months

AI Platform Engineer

Wolters Kluwer

Architected a Medical Research AI Platform modernizing a legacy system, enabling researchers to build search strategies, interact with document and journal chatbots, run AI-driven searches, and generate automated summaries.
Handled end-to-end design and implementation of a centralized MCP server that auto-generated tool capabilities from API metadata and supported service-level filtering for agents, reducing new API onboarding effort by 85%, enabling 200+ enterprise tools through a unified interface, improving tool invocation success rates by 28%, and decreasing orchestration latency by 42%.
Led adoption of Specification-Driven Development (SDD) across enterprise AI and platform projects, establishing specification-first workflows, automated scaffolding, and AI-assisted implementation patterns that reduced feature development time by 40%, decreased requirement-to-production cycles by 30%, and improved engineering productivity across multiple teams.
Developed a multi-agent system using LangGraph and AWS Bedrock AgentCore to orchestrate concurrent agents, reducing end-to-end response latency by 38%, lowering inference cost by 31%, and improving task success rate by 22%.
Designed a RAG pipeline that ingested clinical documents with LlamaParse, applied k-means clustering, stored data in PostgreSQL databases, and used hybrid search with custom embeddings.
Architected a GraphRAG-based clinical intelligence platform, processing millions of medical documents through LlamaParse, constructing domain-specific knowledge graphs from extracted entities and relationships, and integrating PostgreSQL storage layers.
Implemented graph-aware retrieval, custom embedding models, and multi-hop reasoning workflows, increasing retrieval precision by 30% and reducing hallucinations by 35% for complex medical research and evidence discovery use cases.
Evaluated retrieval and generation quality using RAGAS metrics including faithfulness, answer relevance, context precision, and context recall, improving overall RAG quality by 26%, increasing faithfulness by 19%, and reducing hallucinations by 34%.
Processed and transformed large-scale clinical datasets in Databricks using PySpark, improving ETL throughput by 4.2x, reducing batch processing time by 57%, and increasing pipeline reliability by 29% across distributed workloads.
Fine-tuned Meditron-7B on AWS SageMaker to power a medical reasoning agent, leveraging PubMed articles, abstracts, and medical guidelines to improve step-by-step clinical reasoning and answer reliability.
Applied prompt engineering across agents to improve answer accuracy by 21%, reduce hallucinations by 28%, and increase consistency in multi-step reasoning workflows.
Developed and fine-tuned HIPAA-compliant guardrail models using BERT to enforce privacy, compliance, and safety constraints.
Built a human-in-the-loop guardrail monitoring pipeline to detect unsafe outputs, alert clinicians, and track metrics such as violation frequency, confidence thresholds, false-positive rate, and escalation rate, reducing policy violations by 41% and false positives by 24%.
Optimized LLM inference with vLLM, benchmarking against ONNX Runtime and TensorRT to support low-latency, high-concurrency multi-agent processing, improving throughput by 33% and reducing p95 latency by 27%.
Built full MLOps pipelines on AWS using Terraform, Lambda, ECS, EKS, SageMaker, CloudWatch, ECR, S3, API Gateway, and VPC Link, with Docker, Kubernetes, and Bitbucket for deployment, monitoring, and automated retraining.
Developed monitoring and evaluation frameworks with MLflow to track each agent, prompt, retrieval step, and model version, improving observability, auditability, and experiment traceability beyond CloudWatch.

Apr 2021 - May 2024

3 years 2 months

Senior AI Engineer

Sage

Developed a multi-agent crypto analytics platform that combined market forecasting, trading signals, and social sentiment analysis to generate actionable portfolio strategies, improving recommendation quality by 12% and increasing user engagement by 15%.
Built prediction and signal-generation agents using Python, PyTorch, TensorFlow, and NLP models, improving market trend prediction accuracy by 8% and achieving 80–85% sentiment classification accuracy across Twitter and Reddit datasets.
Designed a recommendation engine that aggregated outputs from multiple AI agents into risk-adjusted portfolio strategies, reducing portfolio evaluation time by 35% and improving recommendation relevance by 10%.
Implemented end-to-end batch and real-time data pipelines for portfolio, market, and social data using GCP Cloud Storage and Pub/Sub, reducing data processing latency by 30% and supporting the ingestion of 100K+ events daily.
Deployed scalable agent services on GCP using Google Kubernetes Engine (GKE) and Cloud Functions, supporting 3x–5x workload growth while maintaining 99.5% system availability.
Managed model versioning, experiment tracking, and reproducible ML workflows with MLflow, reducing model deployment cycles by 20% and accelerating experimentation velocity by 25%.
Developed interactive dashboards and portfolio analytics visualizations, reducing user analysis effort by 25% and increasing actionable insight adoption by 15%.
Orchestrated multi-agent workflows and recommendation pipelines, reducing end-to-end recommendation generation time by 20–30% while ensuring consistent and reliable outputs.
Implemented monitoring, logging, and alerting with Stackdriver, reducing incident detection time by 30% and improving operational reliability to 99.5% uptime.

Jan 2019 - Mar 2021

2 years 3 months

Bucharest, Romania

Data Scientist

Amazon

Designed and validated machine learning models for personalized product recommendations and ranking, partnering with engineers to productionize offline models into low-latency SageMaker services, improving CTR by 4–6% and user engagement by 8–10%.
Built and maintained end-to-end data pipelines using AWS S3, AWS Glue, Lambda, and EMR to ingest, clean, and transform large-scale clickstream, product metadata, and review datasets for model training and analysis.
Developed NLP workflows to extract signals from customer reviews, search queries, and support tickets, including tokenization, embedding generation, and intent classification, helping improve product search and catalog understanding.
Fine-tuned BERT and other transformer-based models, alongside classical ML models such as logistic regression and gradient-boosted trees, to improve query understanding, sentiment analysis, and cold-start recommendations.
Applied feature engineering and feature store practices with versioned features and lineage tracking to improve experiment reproducibility and reduce model iteration time by 20–25%.
Built SageMaker training and evaluation workflows, including hyperparameter tuning and model packaging, reducing turnaround time from experiment to production by 20–30%.
Converted offline ranking experiments into production inference workflows on SageMaker and Spark, improving CTR by 4–6% and reducing serving latency by 25–35% through optimized feature pipelines and lighter model artifacts.

Aug 2016 - Dec 2018

2 years 5 months

Bucharest, Romania

Software Engineer

Zitec

Developed scalable data ingestion and ETL pipelines to consolidate CRM, transactional, and marketing data from multiple sources, improving data availability for downstream reporting and reducing manual processing time by 25%.
Engineered customer segmentation and predictive analytics workflows using Python, SQL, and scikit-learn, enabling more targeted campaigns and improving marketing ROI by 10–12%.
Built churn and conversion prediction services using statistical models and regression techniques, helping optimize targeting strategies and increasing customer retention by 6–8%.
Created automated dashboards and reporting pipelines using SQL and Tableau/BI tools, enabling near real-time visibility into key metrics for internal teams and external stakeholders.
Developed reusable Python and SQL modules for data preprocessing, feature extraction, and report generation, reducing processing time by 15–20% and improving code maintainability.
Integrated analytics tracking and event logging frameworks to improve data quality, support cross-platform attribution, and increase reporting accuracy by 10%.

Industry Experience

See where this freelancer has spent most of their professional time.

Experienced in Information Technology, Banking and Finance, Healthcare, Pharmaceutical, and Retail.

Information Technology

Banking and Finance

Healthcare

Pharmaceutical

Retail

Business Area Experience

See which departments and functions this freelancer has contributed to most.

Experienced in Information Technology, Business Intelligence, Research and Development, Finance, and Investments and M&A.

Information Technology

Business Intelligence

Research and Development

Finance

Investments and M&A

Summary

Innovative AI Engineer with 10 years of experience designing, deploying, and scaling production-grade AI systems across healthcare, fintech, and e-commerce. Expertise in generative AI, including multi-agent systems, agentic RAG, LLM fine-tuning, inference optimization, and MCP integration, with deep end-to-end ownership of ML/LLMOps pipelines from data preparation and model training to CI/CD, observability, compliance, and security. Proven ability to deliver secure, scalable enterprise AI solutions on AWS, Azure, and GCP while partnering closely with product, data, and infrastructure teams to accelerate delivery, improve operational efficiency, and drive business growth.

Skills

Programming Languages: Python, Typescript, Javascript, Java, Sql, R, C/C++
Backend & Frontend: Fastapi, Flask, Django, Node.Js, Express, React, Next.Js, Tailwind Css, Rest Apis, Graphql, Grpc, Microservices Architecture, Real-Time Systems (Websockets, Sse), Authentication & Authorization (Oauth2, Jwt), State Management (React Query, Swr), Webpack, Babel, Vercel, Jest
Data Engineering & Databases: Postgresql, Mysql, Mongodb, Elasticsearch, Hadoop, Pyspark, Databricks (Lakehouse, Delta Lake), Vector Databases (Faiss, Pinecone, Chroma, Pgvector)
Machine Learning & Deep Learning: Pytorch, Tensorflow, Keras, Scikit-Learn, Numpy, Pandas, Matplotlib
Natural Language Processing & Llm Systems: Gpt, Claude, Gemini, Llama, Grok, Prompt Engineering, Fine-Tuning (Lora / Qlora), Retrieval-Augmented Generation (Rag), Semantic Search, Conversational Ai, Dialogue Systems, Agentic Ai Systems, Voice Ai
Llm Inference & Optimization: Vllm (High-Throughput Inference), Onnx / Onnx Runtime, Tensorrt / Tensorrt-Llm, Quantization (Int8, Fp16, Gptq, Awq), Model Serving Optimization (Batching, Kv Cache, Speculative Decoding), Gpu Optimization (Cuda, Multi-Gpu, Distributed Inference)
Computer Vision & Multimodal Ai: Ocr (Tesseract, Cloud Vision Apis), Object Detection (Yolo), Semantic Segmentation, Multimodal Architectures, Diffusion Models
Cloud Ai & Managed Services: Google Cloud Platform (Gcp), Vertex Ai (Training, Pipelines, Matching Engine), Azure Ai (Azure Openai, Cognitive Services, Azure Ml), Aws Ai Services (Sagemaker, Bedrock, Rekognition, Textract, Comprehend), Serverless Ai (Aws Lambda, Azure Functions, Cloud Functions), Cloud Ocr (Textract, Azure Ocr, Google Vision Api), Cloud Storage & Data (S3, Gcs, Azure Blob Storage, Bigquery, Redshift)
Mlops & Ai Infrastructure: Docker, Kubernetes, Terraform, Ansible, Ci/Cd (Github Actions, Jenkins), Airflow, Kubeflow, Mlflow, Weights & Biases, Prometheus, Grafana, Model Deployment & Monitoring, Hugging Face Ecosystem
Llm Frameworks & Data Processing: Langchain, Langgraph, Langsmith, Llamaindex, Llamaparse, Autogen, Crewai, Dify, Livekit (Realtime Ai), Model Context Protocol (Mcp)

Languages

English

Advanced

Romanian

Advanced

Education

May 2014 - Jun 2016

University of Bucharest

Master of Computer Science · Computer Science · Bucharest, Romania

Sep 2010 - Apr 2014

University of Bucharest

Bachelor of Computer Science · Computer Science · Bucharest, Romania

Statistics

Experience

Total positions 4

Experience in Information Technology 5.5 y

Avg length 2 y 4 m

Longest experience 3 y 1 m

Global Experience

Countries worked in 1 (Romania)

Primary country Romania

Expertise

Recent roles AI Platform Engineer, Senior AI Engineer, Data Scientist

Main industries Information Technology, Banking and Finance, Healthcare

Main business areas Information Technology, Business Intelligence, Research and Development

Qualifications

Highest degree Master

Profile

Created

July 2026

Need a freelancer? Find your match in seconds.

Try FRATCH GPT

Frequently asked questions

Have questions? Find more information here.

Robert is based in Bucharest, Romania.

Robert speaks the following languages: English (Advanced), Romanian (Advanced).

Robert has at least 10 years of experience. During this time, Robert has worked in at least 4 different roles and for 4 different companies. The average length of individual experience is 2 years and 5 months. Note that Robert may not have shared all experience and actually has more experience.

Based on recent experience, Robert would be well-suited for roles such as: AI Platform Engineer, Senior AI Engineer, Data Scientist.

Robert's most recent position is AI Platform Engineer at Wolters Kluwer.

In recent years, Robert has worked for Wolters Kluwer and Sage.

Robert is most experienced in industries like Information Technology, Banking and Finance, and Retail. Robert also has some experience in Healthcare and Pharmaceutical.

Robert is most experienced in business areas like Information Technology, Business Intelligence, and Research and Development. Robert also has some experience in Finance and Investments and M&A.

Robert has recently worked in industries like Banking and Finance, Information Technology, and Healthcare.

Robert has recently worked in business areas like Information Technology, Business Intelligence, and Finance.

Robert holds a Master in Computer Science from University of Bucharest and a Bachelor in Computer Science from University of Bucharest.

The availability of Robert needs to be confirmed.

Robert's rate depends on the specific project requirements. Please use the Meet button on the profile to schedule a meeting and discuss the details.

To hire Robert, click the Meet button on the profile to request a meeting and discuss your project needs.