Looking for a job

Nare Torosyan

@naretorosyan

Machine learning researcher at Ki Insurance who improved risk predictions on unreliable data across five independent datasets.

Armenia

Message

At Ki Insurance, I developed a confidence-aware machine learning method for insurance risk predictions with unreliable input data. It lowered deployment deviance, with bootstrap confidence intervals excluding zero across six comparisons, and replicated across five independent datasets.

I also built the experimental pipeline, including a synthetic policy generator, configurable corruption engine, and evaluation framework. I extended the Marginalizing Corrupted Features framework and compared several methods against baselines, delivering the code through reviewed PRs in Ki’s GitHub on Google Cloud.

At Data42, I designed and maintained Python and PostgreSQL ETL pipelines orchestrated with AWS Step Functions. I also built an audio-to-insight pipeline using AWS Transcribe and LLM summarisation, migrated models to dbt, and created Apache Superset dashboards.

Experience

Work history, roles, and key accomplishments

Machine Learning Researcher

Ki Insurance

Jun 2026 - Sep 2026 (3 months)

Developed a confidence-aware ML method for robust insurance risk predictions and built the full experimental pipeline including synthetic data generation and evaluation frameworks.

Machine Learning PyTorch scikit learn GLM Uncertainty Quantification Statistical Inference Experimental Design Python SQL Google Cloud Platform Git

Bioinformatics Research Participant

IIAP, National Academy of Sciences of Armenia

Sep 2024 - Dec 2024 (3 months)

Research placement in a molecular dynamics group working with GROMACS simulation tooling and Python scientific-computing workflows.

GROMACS Python Scientific Computing

Data Engineer

Data42

Apr 2023 - Aug 2024 (1 year 4 months)

Designed and maintained Python/PostgreSQL ETL pipelines, built an audio-to-insight pipeline with AWS Transcribe and LLM summarisation, and migrated models to dbt with dashboards in Apache Superset.

Python SQL PostgreSQL ETL AWS Step Functions AWS Lambda S3 AWS Transcribe DBT Apache Superset Git LLM Integrations Prompt Engineering

Software Engineering Intern

Byblos Group

Jan 2023 - Mar 2023 (2 months)

Built a C# desktop application backed by MySQL to collect and analyse application usage data.

C#MySQL desktop application development

Co-founder & Developer

Smarty

Nov 2019 - Oct 2021 (1 year 11 months)

Built an ed-tech application prototype exploring gamification and personalised learning; achieved 2nd place at ICEGOV 3.0 and was a finalist at Microsoft Next AI Guardians.

Edtech Gamification Personalized Learning Prototyping

Data Analyst Intern

Central Bank of the Republic of Armenia

Jun 2021 - Sep 2021 (3 months)

Performed data modelling and reporting on internal banking data using SAP BI tools.

Data Modeling Reporting SAP BI

Education

Degrees, certifications, and relevant coursework

University College London

Master of Science, Data Science and Machine Learning

2025 -

Activities and societies: Chevening Scholar; modules include Applied Deep Learning, Applied Machine Learning, Graphical Models, Statistical Design of Investigations, and Introduction to Statistical Data Science.

MSc in Data Science and Machine Learning, with an industry thesis on confidence-aware machine learning for robust insurance risk modelling.

Université Paul Sabatier Toulouse III

Bachelor of Science, Computer Science

2019 - 2023

Grade: Honours

BSc in Computer Science (Informatics and Applied Mathematics) as part of a double diploma with the French University in Armenia, graduated with Honours.

French University in Armenia

Bachelor of Science, Computer Science

2019 - 2023

Grade: Honours

BSc in Computer Science (Informatics and Applied Mathematics) with a double diploma from Université Paul Sabatier Toulouse III, graduated with Honours.