Saffat Bokul

Winnipeg, Canada0

Machine Learning / AI Engineer | Production ML & Data Systems | Python • PyTorch • AWS | RAG, Agents & ML Systems

I work at the intersection of machine learning, Artificial Intelligence, software engineering, and data infrastructure. I partner with business, strategy, and operations leaders to understand the decision they actually need to make, decompose it into a technical problem, and build the pipeline, model, or agent that makes that decision repeatable. Currently, I’m in Strategy, Planning & Analytics at The North West Company, a $3B multi-banner retailer operating across Canada, Alaska, the Caribbean, and the South Pacific. My work includes: • Building internal agentic AI workflows over governed enterprise data to automate recurring analytical work—producing structured briefings, retrieving metric definitions, surfacing promo/labour exceptions, and helping strategy teams move from raw data to decisions faster. • Engineering some of the company’s first in-house dbt semantic models and Dagster-orchestrated Snowflake pipelines, creating a governed and reusable data foundation for executive and operational reporting across sales, margin, labour, shrink, promotions, out-of-stocks, private label, and other core KPIs. • Developing promotion and labour decision-support analytics across multiple retail banners, using lift-vs-baseline, price sensitivity, promotion effectiveness, UPLH, forecasting, and profitability analysis to identify performance drivers, flag underperforming initiatives, and support better pricing, promotion, and labour-allocation decisions. Previously at Birchwood Automotive Group, I built production data and ML systems across AWS and Snowflake—engineering pipelines processing $1B+ records, helping create a 500K+ customer identity layer, delivering integrations contributing to ~$3M in annual savings, and training/deploying predictive ML models through Amazon SageMaker. My ML foundation comes from my M.Sc. in Computer Science, where I developed deep-learning models for molecular transition-state discovery using Python and PyTorch, reducing a computational process from weeks to hours. I am focusing on Production AI/ML systems: PyTorch, RAG, agentic AI, LangGraph, Hugging Face, vLLM/SGLang, inference/model serving, AWS/SageMaker, Docker, Snowflake/dbt/Dagster, evaluation, observability, and distributed systems. Interested in ML Engineer, AI Engineer, SWE–ML, ML Platform, and AI Platform roles in Toronto/GTA or Vancouver. Open to relocation.

AI Fluency

No AI stories yet

Nothing posted yet.

Experience

Data Analyst

The North West Company

Mar 2026 — Present · Winnipeg, Manitoba, Canada

• Prototyping agentic AI workflows that combine enterprise data access, business context, and tool-based analysis to automate repeatable analytical investigations and produce structured, traceable decision support. • Developed Bayesian ML, causal inference, synthetic-control, and price-elasticity models across different product classes to measure true promotional incrementality and discount sensitivity, identifying “discount cliffs,” separating genuine demand lift from margin-destructive promotions, and surfacing opportunities to improve sales, volume, and margin performance. • Building the data foundation for an enterprise executive performance platform designed for daily use by senior leadership, engineering Snowflake data pipelines, reusable dbt transformation models, and Dagster-orchestrated workflows in a Linux environment to feed governed Tableau reporting. • Own key technical and analytical work within enterprise promotion-performance initiatives spanning retail operations in Canada, Alaska, and the Caribbean, developing reproducible analyses for incremental sales/volume/margin lift, promotional baselines, price sensitivity, promotion effectiveness, and product/promotion interactions.

Data Analyst

Birchwood Automotive Group

Sep 2024 — Feb 2026 · Winnipeg, Manitoba, Canada

Built the data and ML platform behind retail analytics for a large automotive group. Official title was Data Analyst; the work spanned data engineering, orchestration, APIs, ML development, model serving, and data science. Data & ML Platform: • Built Snowflake dimensional models with 100B+ rows using Python, PySpark, dbt, and REST/SOAP APIs, improving operational query performance by ~30% and establishing reusable, tested transformation layers for analytics and ML. • Engineered Dagster/Airflow pipelines with CI/CD and Grafana monitoring across jobs and Snowflake usage, reducing failed runs by ~90% and improving pipeline observability and reliability. • Built S3 + Snowpipe ingestion for dealer-management data and FastAPI services over curated datasets, enabling 3+ teams to self-serve data while replacing third-party reporting/consulting costs worth ~$3.3M annually. ML / AI in Production • Built an entity-resolution pipeline using Python, EMR, S3, and EC2 across 20M+ records / 10+ sources, creating a 500K+ unified customer view and reducing duplicate identity noise by ~95%. • Developed versioned feature pipelines and a foundational Snowflake customer asset supporting 5+ ML/analytics capabilities including churn, CLV, segmentation, targeting, and retention. • Trained, evaluated, deployed, and monitored churn/survival models in Amazon SageMaker, surfacing predictions through Sigma to 500+ users and supporting retention strategies that reduced churn by ~25% in targeted segments. • Applied Amazon Bedrock to unstructured customer-feedback scoring and causal/marketing-mix modeling to ~$250K+ in spend, contributing to ~5-point NPS improvement and ~30% lower CAC through better targeting and budget allocation.

Graduate Research Assistant

University of Manitoba

Sep 2021 — Oct 2023 · Winnipeg, Manitoba, Canada

• Accelerated the process of finding transition states (for drug design) from a few weeks to hours by theorizing and implementing a novel machine learning architecture leading to novel candidate transition states. • Processed data of 100,000 complex molecules in the form of graphs using Python, PyTorch, Sci-kit Learn and created visualizations, dashboards and reports for data analysis. • Examined and tuned experimental deep learning models using a series of metrics to clearly communicate findings with stakeholders, to plan future research direction. • Collaborated with interdisciplinary researchers to explore the use of artificial intelligence in pharmacology, self-learning about techniques like computational chemistry, molecular dynamics, graph machine learning in 3 months.

Software Engineer - QA

Orbitax

Aug 2020 — Jul 2021 · Dhaka, Bangladesh

• Developed test automation suite using JMeter saving 100 hours each week in combined QA time, by using an agile approach to test automation which delivered value and minimized technical debt. • Acquired the ownership of testing the Orbitax Entity Tracker and resolved over 100 bugs by collaborating with the product manager, developers, and other stakeholders. • Identified test scenarios by reviewing user stories for Orbitax International Tax Platform, and created test cases, designed test scopes, set testing strategy with 90% coverage. • Helped with onboarding and mentoring new team members, created training guides, offered one-on-one mentorship leading to 50% quicker onboarding.