Daniel is available for hire
Hire DanielDaniel Polimac
Verified Expert in Engineering
Artificial Intelligence Engineer and Developer
Belgrade, Serbia
Toptal member since August 17, 2026
Bio
Daniel is an intelligent systems engineer specializing in reliable AI, signal processing, and edge-to-cloud inference. He builds systems that turn noisy, incomplete, or complex real-world data into trustworthy knowledge and actions. His work spans production AI, physiological and sensor-signal processing, speech and computer vision, embedded inference, knowledge-grounded systems, and AI infrastructure.
Portfolio
Roboxi
PyTorch, NVIDIA CUDA, OpenCV, Terraform, Microsoft Azure, Python, Docker Swarm...
Future Ready AS
PyTorch, Hugging Face Transformers, FastAPI, NVIDIA CUDA, Grafana, Terraform...
Lifeness
PyTorch, Transformers, FastAPI, Amazon Web Services (AWS), Terraform, Python...
Experience
- Machine Learning - 11 years
- Signal Processing - 10 years
- Knowledge Systems - 10 years
- AI Systems Architecture - 10 years
- Machine Learning Operations (MLOps) - 8 years
- Real-Time Inference - 8 years
- Sensor Fusion - 7 years
- Edge AI - 7 years
Preferred Environment
Linux, Amazon Web Services (AWS), Microsoft Azure, NVIDIA CUDA, Docker, PyTorch
The most amazing...
...system I've architected is a patented real-time physiological state classification pipeline deployed on embedded devices.
Work Experience
AI Infrastructure Engineer – Robotics Perception
2025 - 2026
Roboxi
- Designed perception infrastructure for semi-autonomous mobile robotics, integrating sensor fusion and computer vision into low-latency real-time pipelines.
- Integrated model inference, sensor processing, and robotics software for dependable operation in dynamic, unstructured environments.
- Fine-tuned Vision Transformer models to detect foreign object debris, surface cracks, and other airport-surface anomalies from visual inspection data.
- Applied model quantization to reduce inference cost and latency while preserving detection quality for real-time and near-real-time inspection workloads.
- Implemented dynamic batching in Triton to improve GPU utilization and throughput across concurrent inspection requests.
Technologies: PyTorch, NVIDIA CUDA, OpenCV, Terraform, Microsoft Azure, Python, Docker Swarm, NVIDIA Triton, Real-time Data, MLflow, AI Systems Architecture, Site Reliability Engineering (SRE), Real-Time Inference, Cloud Architecture, Edge AI, Artificial Intelligence (AI), Computer Vision, Machine Learning, Large Language Models (LLMs), Data Engineering, Classification, Optical Character Recognition (OCR), Data Pipelines, Data Extraction, Back-end, FastAPI, Retrieval-augmented Generation (RAG), Vector Databases, Software Engineering
Principal AI Engineer – Speech Systems & MLOps
2024 - 2026
Future Ready AS
- Architected and shipped a real-time streaming speech-recognition system for conversational AI operating on noisy, uncontrolled audio.
- Built a GPU-accelerated back end for bidirectional audio/text streaming and optimized endpointing, buffering, scheduling, and inference behavior for production latency and stability.
- Fine-tuned Whisper models for Norwegian speech and adapted recognition for Bokmål and Nynorsk language requirements across target speakers and acoustic environments.
- Re-engineered Whisper from fixed around 30-second inference windows into real-time streaming transcription with incremental partial results for conversational use cases.
- Built a high-throughput batch transcription pipeline for call-center QA, enabling recorded conversations to be processed and analyzed automatically at scale.
- Extended the speech platform into personalized AI tutor and agent workflows using transcripts and conversational context to support individualized interactions.
Technologies: PyTorch, Hugging Face Transformers, FastAPI, NVIDIA CUDA, Grafana, Terraform, TypeScript, Python, Amazon Web Services (AWS), Signal Processing, AI Agents, Site Reliability Engineering (SRE), Real-Time Inference, Knowledge Systems, Cloud Architecture, Edge AI, Neo4j, Machine Learning, Artificial Intelligence (AI), Large Language Models (LLMs), Data Engineering, Classification, Document Processing, Data Pipelines, Data Extraction, Back-end, Retrieval-augmented Generation (RAG), Vector Databases, Sentiment Analysis, Speech-to-Text (STT), Software Engineering
AI Systems Engineer – LLM Inference & Knowledge Systems
2023 - 2025
Lifeness
- Architected and shipped a production LLM/RAG system for personalized health guidance, combining fine-tuned models with retrieval over an evidence-based knowledge base.
- Designed the grounding layer so that generated responses remained traceable to source material and clinician-authored behavioral and nutrition protocols constrained model output.
- Built a multimodal food-image recognition and nutrient-classification pipeline for low-latency cloud inference.
- Worked with clinicians to encode behavioral science and nutrition protocols as explicit constraints and domain knowledge within the reasoning layer.
- Designed the surrounding production AI stack, including retrieval, model serving, orchestration, and deployment workflows for reliable health-focused AI applications.
- Integrated retrieval, model serving, and domain-knowledge components into a modular production architecture supporting scalable health-focused AI applications.
Technologies: PyTorch, Transformers, FastAPI, Amazon Web Services (AWS), Terraform, Python, PostgreSQL, Vector Stores, AI Agents, AI Systems, Knowledge Systems, Cloud Architecture, Neo4j, Large Language Models (LLMs), Statistics, Computer Vision, Machine Learning, Artificial Intelligence (AI), Data Engineering, Classification, Document Processing, Optical Character Recognition (OCR), Data Pipelines, Data Extraction, Back-end, Retrieval-augmented Generation (RAG), Vector Databases, Speech-to-Text (STT), Software Engineering
Senior Machine Learning Engineer
2020 - 2024
Sleep Number
- Co-developed and productionized a patented real-time physiological state-classification pipeline, spanning biosignal processing, embedded inference, and deployment in commercial sleep-tracking products.
- Architected embedded inference on ARM microcontrollers under strict latency, memory, and power constraints, integrated with cloud analytics and model lifecycle infrastructure.
- Built sensor-fusion and signal-processing pipelines that improved robustness and classification stability on real-world biometric data.
- Integrated on-device inference with cloud analytics and model lifecycle infrastructure to maintain consistent behavior across deployed devices.
- Worked across ML, embedded, hardware, and product teams to resolve system-level trade-offs involving accuracy, latency, robustness, and constrained compute.
Technologies: Python, C, TensorFlow, Open Neural Network Exchange (ONNX), Databricks, C++, Amazon Web Services (AWS), Signal Processing, Data Science, Real-time Data, MLflow, AI Systems, Site Reliability Engineering (SRE), Real-Time Inference, Knowledge Systems, Edge AI, Neo4j, Cloud Architecture, Statistics, Machine Learning, Artificial Intelligence (AI), Large Language Models (LLMs), Data Engineering, Classification, Data Pipelines, Data Extraction, Back-end, Life Science, Software Engineering
Research Engineer
2021 - 2023
Norwegian University of Life Sciences
- Designed and deployed a cloud ML system for non-contact heart-rate estimation using remote photoplethysmography from facial video.
- Developed real-time computer vision and signal-processing pipelines to recover stable physiological signals from subtle facial color and motion changes.
- Engineered cloud execution and signal-quality controls for reproducible physiological measurements across large experimental cohorts.
- Integrated physiological sensing with Qualtrics, linking biometric and psychological data across thousands of study participants.
Technologies: PyTorch, OpenCV, Amazon Web Services (AWS), Python, Signal Processing, Computer Vision, Deep Learning, Data Science, AI Systems, Real-Time Inference, Cloud Architecture, Edge AI, Machine Learning, Large Language Models (LLMs), Data Engineering, Classification, Data Pipelines, Data Extraction, Back-end, Software Engineering
AI Researcher — Biological Sequence Learning & Clinical Risk Modeling
2019 - 2020
Temple University
- Developed machine-learning pipelines for clinical risk modeling using multimodal patient data spanning tabular, temporal, and clinical variables.
- Built predictive models for type 2 diabetes complications to identify high-risk patient cohorts and disease trajectories from heterogeneous clinical data.
- Applied language-inspired sequence embeddings to protein contact prediction, supporting structural and functional inference from biological sequence data.
- Co-authored peer-reviewed research on predicting complications of diabetes mellitus, published in the Journal of the American Medical Informatics Association.
Technologies: Python, PyTorch, Scikit-learn, Multimodal Data Analysis, Clinical Risk Modeling, Bioinformatics, Sequence Modeling, Model Interpretability, Neo4j, Causal Inference, Statistics, Machine Learning, Computational Biology, Computer Vision, Artificial Intelligence (AI), Large Language Models (LLMs), Classification, Optical Character Recognition (OCR), Data Pipelines, Data Extraction, Life Science, Software Engineering
AI Research Engineer – Clinical NLP & Information Extraction
2018 - 2019
German Research Center for Artificial Intelligence GmbH (DFKI)
- Built clinical NLP pipelines for entity recognition, relation extraction, and concept normalization over German reports, combining statistical models with symbolic medical resources.
- Developed preprocessing and concept-normalization workflows to improve the representation of complex German clinical language for downstream extraction.
- Integrated tools, including spaCy, Flair, Apache cTAKES, and MetaMap, into a unified clinical information-extraction workflow.
- Developed research prototypes for transforming unstructured clinical text into structured representations suitable for downstream analysis and knowledge-based reasoning.
Technologies: TensorFlow, Python, Natural Language Processing (NLP), AI Systems, Knowledge Systems, Neo4j, Data Science, Causal Inference, Bioinformatics, Machine Learning, Artificial Intelligence (AI), Data Engineering, Classification, Document Processing, Optical Character Recognition (OCR), Data Pipelines, Data Extraction, Back-end, Life Science, Software Engineering
Lead Data Scientist - Collective Intelligence & Recommendation Systems
2016 - 2018
Enetel Solutions
- Led the design and deployment of a real-time recommendation system for a large-scale eCommerce platform using Spark and Hadoop over high-volume behavioral data.
- Built customer segmentation and personalization models to improve product targeting and support data-driven recommendation workflows.
- Established continuous A/B testing pipelines to evaluate recommendation changes and measure production impact across user segments.
- Managed a small data science team and improved observability, reliability, and maintainability across production data and ML pipelines.
Technologies: Python, Apache Spark, Hadoop, Docker, Business Intelligence (BI), Data Science, TensorFlow, Microsoft Azure, MLflow, AI Systems, Knowledge Systems, Cloud Architecture, Data Engineering, Classification, Document Processing, Optical Character Recognition (OCR), Data Pipelines, Data Extraction, Back-end, Sentiment Analysis, Software Engineering
Experience
AI-powered Airport Surface Inspection
https://www.roboxi.com/I built an end-to-end computer vision pipeline for automated airport surface inspection, covering data ingestion, preprocessing, model training, evaluation, and deployment. I fine-tuned Vision Transformer models to detect foreign object debris, surface cracks, and related runway or apron anomalies from visual inspection data. I then optimized production inference with the NVIDIA Triton Inference Server by applying model quantization and dynamic batching to improve GPU utilization, throughput, and deployment efficiency.
My role covered the full machine learning lifecycle, from data and model development through inference optimization and production serving.
My role covered the full machine learning lifecycle, from data and model development through inference optimization and production serving.
Co-patented an Embedded Sleep-state Classification System
https://www.sleepnumberlabs.com/Developed a real-time physiological state-classification system for commercial sleep-tracking products, taking the solution from noisy biosignal processing through production on-device inference. I built signal processing and sensor-fusion pipelines for robust physiological feature extraction, then architected inference on ARM microcontrollers under strict latency, memory, and power constraints. I also integrated embedded predictions with cloud analytics and model lifecycle infrastructure to maintain consistent behavior across deployed devices.
The resulting technology was patented, and I am a named inventor on
U.S. Patent 12,369,848 B2.
The resulting technology was patented, and I am a named inventor on
U.S. Patent 12,369,848 B2.
Education
2016 - 2020
Doctorate in Intelligent Systems
University of Belgrade - Belgrade, Serbia
2015 - 2016
Master's Degree in Information Systems and Technologies
University of Belgrade - Belgrade, Serbia
2010 - 2015
Bachelor's Degree in Computer Engineering
School of Computing - Belgrade, Serbia
Skills
Libraries/APIs
PyTorch, Hugging Face Transformers, OpenCV, TensorFlow, Scikit-learn
Tools
Grafana, Terraform, Open Neural Network Exchange (ONNX), Docker Swarm
Languages
Python, C, C++, TypeScript
Platforms
Databricks, Docker, Linux, Amazon Web Services (AWS), NVIDIA CUDA
Storage
PostgreSQL, Data Pipelines, Neo4j
Frameworks
MediaPipe, Apache Spark, Hadoop
Paradigms
Business Intelligence (BI)
Industry Expertise
Bioinformatics
Other
FastAPI, Transformers, Site Reliability Engineering (SRE), Real-time Data, Edge AI, NVIDIA TensorRT, AI Agents, Signal Processing, Computer Vision, Deep Learning, Data Science, Natural Language Processing (NLP), MLflow, Machine Learning, Software Development, Information Systems, AI Systems Architecture, Knowledge Systems, Real-Time Inference, Sensor Fusion, Machine Learning Operations (MLOps), Cloud Architecture, AI Systems, Artificial Intelligence (AI), Large Language Models (LLMs), Causal Inference, Statistics, Data Engineering, Classification, Document Processing, Optical Character Recognition (OCR), Data Extraction, Back-end, Sentiment Analysis, Speech-to-Text (STT), Software Engineering, Microsoft Azure, Computational Biology, Retrieval-augmented Generation (RAG), Vector Databases, Life Science, NVIDIA Triton, Vector Stores, Hardware Development, Multimodal Data Analysis, Clinical Risk Modeling, Sequence Modeling, Model Interpretability, Knowledge Management
Collaboration That Works
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
1
Share your needs
Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2
Choose your talent
Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3
Start your risk-free talent trial
Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.
Top talent is in high demand.
Start hiring