Nikiforos Mandilaras

Nikiforos Mandilaras

Senior ML/AI Engineer

👋 About Me

I’m a Senior Machine Learning Engineer who builds production-grade ML systems end to end. With a track record at major companies across diverse business sectors, I work at the intersection of modeling, engineering, and real-world constraints, turning ambiguous problems into deployed solutions. I’m most energized by shipping systems that create measurable impact, not just training models.

🧩 Execution & Philosophy

Ownership & Agency: I take responsibility for outcomes. From data sourcing to production deployment, I assume problems are solvable and work through obstacles to deliver critical results.

Solve the real problem: I focus on business impact, not just task completion. If a system is inefficient, I’d rather redesign it than patch it.

End-to-end thinking: I’m comfortable working across the stack, from debugging low-level code to designing LLM pipelines.

Bias toward building: I prototype early, test in real conditions, and iterate fast. Working systems beat perfect plans.

Clear communication: I keep stakeholders aligned with concise updates focused on results and next steps, reducing the need for management overhead.

🛠️ Work Experience

Taboola Nov 2021 – Present

Senior Machine Learning Engineer - Tech Lead

  • Working with state-of-the-art algorithms and technologies in a large-scale Recommender System.
  • Applying LLMs for representation learning, leveraging fine-tuned embeddings in multiple downstream tasks.
  • Delivered an incremental learning framework with 12x faster training and deployment cycles, achieving revenue lift.
  • Developed a new candidate generation model using Vespa, leveraging a two-tower architecture to efficiently provide online ranking of suitable items.
  • Part of the algo expert team, leading high-priority initiatives, designing workflows, and providing guidance and code reviews for the rest of algo department.
Ubitech June 2021 – Nov 2021

Machine Learning Engineer

  • Designed ML solutions for smart-grid use cases such as self-consumption optimization, energy savings estimation, and flexibility prediction.
  • Worked with research and engineering teams to turn ideas into working prototypes for the energy domain.
Intracom Telecom May 2018 – May 2020

Data Scientist / Big Data Engineer

  • Designed and implemented ML features for telco products, including load forecasting, grid optimisation, and resource allocation.
  • Built Big Data applications for telco and banking clients on Hadoop, including high-throughput ingestion pipelines and complex data analysis.

🎓 Education

  • National Technical University of Athens
    • 📘 MSc. in Data Science and Machine Learning, 9.61/10 (Oct. 2018 – Aug. 2020)
      Master Thesis in Resource Allocation with Deep Reinforcement Learning Methods.
    • 📗 Integrated Master in Electrical and Computer Engineering, 8.44/10 (Sept. 2012 – Feb. 2018)
      Major in Computer Science: Software, Computer Systems & Networks.
      Master Thesis in Big Data.

💻 Technical Skills

  • Languages: Python, Java
  • Deep Learning: PyTorch, TensorFlow, Keras, LLMs
  • ML & Big Data: Apache Spark, Kafka, Hadoop, Sklearn, Pandas, Numpy
  • Databases: BigQuery, Vespa, Elastic
  • MLOps: Airflow, Jenkins, Grafana, K8s, Docker
  • Agents: Claude Code, Cursor

🚀 Selected Projects

  • ⚡ LLM-Based Embedding Pipeline:
    Designed and built a distributed production pipeline that generates item embeddings for ad recommendations. Used LLMs for description enrichment and embedding generation. Powers downstream recommendation models with continuously updated item representations.

  • 📈 Incremental CTR Learning Framework:
    Reduced training and deployment overhead by 90% through an incremental learning framework for large-scale CTR prediction that fine-tunes on real-time data streams.

  • 🔍 Self-Supervised Recommendation Signals:
    Unlocked revenue from cold-start and low-signal user segments by architecting a self-supervised embedding framework for improved recommendation accuracy.

  • 🤖 Deep Reinforcement Learning for Caching:
    Improved best-effort application speeds by 4x while maintaining strict performance SLAs by engineering a Deep RL agent managing last-level cache sharing in multi-core systems.

  • 📊 Customer Satisfaction Indexes for Telco:
    Delivered KPIs for millions of telco subscribers using Apache Spark, handling noisy, complex data at scale.

Misc

📄 Publications

  • P. Katsileros, N. Mandilaras et al.
    “An Incremental Learning framework for Large-scale CTR Prediction.”
    Presented at the 16th ACM Conference on Recommender Systems, RecSys ‘22, Seattle, WA, USA.

🏆 Honors & Awards

  • N. Kritikos Scholarship: Scholarship granted for best overall math performance. NTUA - ECE School.
  • C. Papakyriakopoulos Math Award: Distinction for performance in Mathematics during my studies. NTUA - ECE School.

🌍 Languages

  • 🇬🇧 English: Michigan Proficiency (C2)
  • 🇫🇷 French: Sorbonne (B2)

📝 Additional Info

  • Military Service: Fulfilled (Sept. 2020 - June 2021)
  • Erasmus+: Participation in youth exchange projects

✉️ Contact

Feel free to drop me a message below! Alternatively, you can connect with me on LinkedIn or reach me at nikmand@outlook.com.