Senior Data Scientist · Toronto, Canada

Abbas Meghani

I build the ML and data systems behind fraud detection at enterprise scale, from feature pipelines to model inputs, backed by an MSc in Data Science and published ML research.

5+ years building production data & ML systems
15+ enterprise clients onboarded to ML fraud detection
20% lower data extraction latency across deployed models
4 peer-reviewed publications (journal + IEEE)
Portrait of Abbas Meghani

About

Data science that ships to production.

I'm a Senior Data Scientist at Shift Technology, where I work on the ML and data systems that power fraud detection for insurers around the world. My work sits where modelling meets engineering: feature pipelines, entity resolution, and the data layers that keep models fast and reliable for every client.

I hold an MSc in Data Science from HEC Montréal and have published ML research in the Journal of Economic Surveys, using topic models to map thirty years of finance literature. Before data science, I was a backend software engineer, which is why I care as much about latency and reliability as I do about model quality.

Outside work, you'll find me on long hikes, reading, or over a warm cup of tea.

ML & Data

  • Python
  • pandas
  • NumPy
  • scikit-learn
  • R
  • SQL / T-SQL
  • ElasticSearch

Retrieval & GenAI

  • RAG pipelines
  • Embeddings
  • pgvector

Engineering

  • FastAPI
  • C# / .NET Core
  • REST APIs
  • Docker
  • PostgreSQL
  • GitHub Actions

Experience

Where I've worked

  1. Shift Technology

    AI-driven fraud detection for insurance · Feb 2022 – Present

    Senior Data Scientist

    Apr 2024 – Present
    • Optimized ML fraud detection pipelines, reducing data extraction latency by 20% across all deployed model environments.
    • Integrated ElasticSearch entity resolution into the ML inference pipeline, improving model input data quality at query time.

    Data Scientist

    Feb 2022 – Apr 2024
    • Built a factory-based feature preparation library for ML microservices, onboarding 15+ enterprise clients to ML-powered fraud detection.
    • Refactored the ML feature pipeline, saving thousands of compute hours and delivering a 5% company-wide efficiency improvement.
    • Wrote hash-based deduplication filters to ingest tables of up to 9 billion highly duplicated rows within server memory limits.
    • Python
    • C#
    • SQL
    • ElasticSearch
    • Docker
  2. DocuFire

    Document automation · Jun 2021 – Feb 2022

    Software Engineer, Backend

    Jun 2021 – Feb 2022
    • Increased document generation throughput by 15% with async parallel page processing in the core ERP system's PDF module.
    • Migrated the legacy data access layer to a REST API with full unit test coverage, preserving every existing endpoint.
    • C#
    • .NET Core
    • SQL
  3. Research roles

    Part time, alongside my MSc and early career · 2020 – 2023

    Research Assistant, ML Research · HEC Montréal

    May 2021 – Oct 2023
    • Built a Python ML model simulating sports game outcomes, served inside a live ASP.NET Core web app used by the Sports Analytics course.

    Research Consultant · Sentometrics

    Aug 2021 – Oct 2021
    • Scraped, cleaned and annotated a corpus of 33 finance journals, then fit and compared Structural Topic Models in R. This work became the Journal of Economic Surveys paper.

    Research Assistant · ERPsim Lab, HEC Montréal

    Sep 2020 – Apr 2021
    • Validated simulated datasets in Python and wrote SAS guides for students and teachers on the Cortex data science learning platform.

Projects

Things I've built

Featured · 2026

ml-serving-api

A containerized model and retrieval serving service. FastAPI serves a scikit-learn model with per-request latency logging, Postgres persistence and pgvector semantic search, tested with pytest and shipped through GitHub Actions CI.

  • Python
  • FastAPI
  • Docker
  • PostgreSQL
  • pgvector
View on GitHub
$ measured latency, p95
/predict   2.9 ms
/search    32.6 ms

# ~200 requests each, single worker

2021 · Python

Complex Networks of Bike Sharing

Overlaid bike-share usage on the city road network, turned demand into time series and visibility graphs, and analyzed them with network descriptors and community detection.

Read the report

2020 · Python, Tkinter

Convex Hull Visualizer

A GUI that animates Brute Force, Jarvis' March, Graham Scan and Divide & Conquer, implemented from the pseudocode in the original papers.

See the code

2019 · Python

Recyclable Trash Classification

Compared one-vs-many SVMs against CNNs (VGG, ResNet34), tuning depth, pooling, batch norm and data augmentation.

View the poster

Research

Publications

Journal of Economic Surveys · 2023

Thirty years of academic finance

Applied an unsupervised Bayesian model, the Structural Topic Model, together with text mining to map how three decades of finance research have evolved across 33 journals.

Read the paper

Education

Academic background

2019 – 2021

MSc, Data Science

HEC Montréal, University of Montréal

GPA 3.6 / 4.3. Received a scholarship for Fin-ML research, and was hired three times for programming projects, two of which became course capstones.

Machine Learning, Deep Learning, Statistical Learning, Forecasting, Optimization, Complex Networks, Data Mining

2014 – 2018

BEng, Computer Engineering

Rizvi College of Engineering, University of Mumbai

GPA 8.4 / 10, top 5%. One of two teams in my cohort to receive research funding for the capstone project.

Data Structures, OOP, Databases, Distributed Systems, Operating Systems, Applied Mathematics

Contact

Let's talk.

I'm happy to chat about data science roles, ML systems, or research. The fastest way to reach me is email.

[email protected]