About Jovan
Machine Learning Engineer and Backend Developer with more than 10 years of experience delivering AI products from prototype to production. Built and operated end-to-end machine learning systems serving 24 enterprise brands on a platform that processes over 2 trillion events per year, including a Transformer-based model that increased conversion rates by 17%. Combines deep modeling expertise with strong backend engineering across microservices, Apache Kafka, and cloud infrastructure on Google Cloud and AWS. Led a team of 3 engineers, designed experimentation frameworks, and partnered with product and business stakeholders to turn models into measurable revenue.
English
Fluent
Remote only
Primarily works remotely
Experience
- WunderkindStaff Machine Learning EngineerDIGITAL AND ITAugust 2024 - June 2026 (1 year and 10 months)Obrenovac, SerbiaBuilt a Transformer-based send-time model with a custom ranking loss and hard-example mining for rare conversions — lifting conversion rates 17%, migrating 5 enterprise clients off the legacy system, now serving ~430,000 sends daily.Delivered models, MLOps pipelines, and runtime services for 24 brands on infrastructure processing 2+ trillion events/year, scaling 5 brands to fully AI-driven delivery.Established the experimentation and delivery framework (A/B testing, staged rollout, monitoring, exploration-exploitation) that moved 9 successive timing-model generations into production.Engineered a high-fidelity simulator of production marketing workflows — modeling clicks, conversions, and unsubscribes from live data — for offline evaluation of timing and channel models.Deployed a budget-constrained channel optimization model using per-user scoring and dynamic thresholding to steer limited SMS budgets toward the highest-lift cohorts.Architected a campaign prioritization engine, validated it offline, and presented the design to 200+ people to align engineering, product, and business.Led 3 ML engineers across 4 initiatives (timing, simulation, prioritization, channel optimization), partnering with Product, Analytics, and Customer Success.Stack: Google Cloud (Kubernetes, Kafka, Deephaven, Iceberg, Trino, BigQuery), GitLab + Kargo CI/CD. Built the team's first Deephaven unit-testing framework, since adopted by other teams.Replaced a legacy Dataflow identity-resolution pipeline with a lightweight Python FastStream service, cutting maintenance and adding Grafana monitoring.Prototyped internal developer tooling with Claude and other LLMs, including agentic workflows that automate a pre-commit lint-fixing loop.
- Sentinel Labs LimitedSenior Backend DeveloperApril 2024 - December 2025 (1 year and 8 months)• • Owned the full technical lifecycle of 2 products from concept to production as the sole technical leader, covering cloud solution architecture, software architecture, product management, development, and DevOps.• • Built a PCI card issuing service that deducted cryptocurrency from user wallets during fiat transactions, implemented as NestJS and TypeScript microservices communicating through Apache Kafka.• • Launched an API-as-a-service offering for AI agent hosting with a dynamic billing system that charged cryptocurrency per API call through a custom NestJS proxy and a separate transaction processing service.• • Administered service discovery, workload orchestration, and secrets management with HashiCorp Consul, Nomad, and Vault, and developed an AWS Lambda provisioner that automatically created and managed client instances.• • Integrated Web3 and non-Web3 components through secure interface points and maintained rapid, reliable releases with GitHub CI/CD pipelines.
- Noveta Corp (Contract)Natural Language Processing EngineerJanuary 2023 - March 2025 (2 years and 2 months)• • Designed an NLP dialog system in TensorFlow 2.0 that enables non-technical users to create complex, structured queries.• • Achieved an F1 score above 95% across 6 natural language understanding tasks on a limited dataset, near state-of-the-art performance, by fine-tuning BERT with multi-task LSTM heads, multiple input streams, and self-atention.• • Improved model performance further through data augmentation, custom loss functions, conditional random fields, and hyperparameter search with Keras Tuner.• • Coordinated dataset design, collection, cleaning, and annotation with 6 annotators, and released an open-source tool for rapid, hotkey-based multi-label annotation.• • Mentored team members on machine learning, TensorFlow, AI architectures, Git, and Docker.
Recommendations
Be the first to recommend Jovan
Help this freelancer shine by sharing your experience working together.
These freelancer profiles also match your criteria
Agatha Frydrych
Backend Java Software Engineer
4.7
(3)
2
Baptiste Duhen
Fullstack developer
4.6
(4)
5
Amed Hamou
Senior Lead Developer
4
(2)
7
Audrey Champion
Web developer
4.3
(3)
4
Education
- Bachelor of ScienceSingidunum University2016Bachelor of Science