About Me
I'm a Data Science Manager at Gildan, where I lead a team turning data into the AI and decision systems behind a global manufacturer's operations, spanning demand forecasting, factory-floor optimization, and the enterprise AI agents that put those insights in people's hands.
What started as an engineering curiosity about how complex systems work has grown into a career focused on turning messy, real-world data into tools people can actually rely on. Over the years I've worked across telecom, agriculture, and manufacturing, on everything from satellite imagery and climate models to factory-floor sensors, using classical machine learning, deep learning, optimization, and modern AI.
Today my focus is the generative-AI and agentic systems that build on top of that foundation. But I still care most about what drew me in on day one: making complex systems legible, and turning data into decisions people can trust.
M.Sc. Information Technology
University of DhakaBangladesh Jan. 2017 - Oct. 2018
B.Sc. Electrical & Electronic Engineering
Khulna University of Engineering & TechnologyBangladesh Apr. 2010 - Nov. 2014
What Sets Me Apart
Most data scientists pick one lane. My edge is depth and range: a decade of hands-on machine learning and deep learning, now extended into AI, delivered at a scale and across domains that rarely overlap, from the factory floor of a global NYSE-listed manufacturer to satellite and climate data spanning an entire subcontinent.
Foundations, Then Frontier
The AI work isn't where I started; it's what I've grown into. A decade of hands-on machine learning and deep learning (forecasting, optimization, computer vision, graph neural networks), now extended into building enterprise generative AI and agentic systems on top of that strong technical foundation.
On the Factory Floor
Inside one of the world's largest apparel manufacturers, the machines never stop. I put machine learning directly into the manufacturing process (spinning, knitting, dyeing, sewing, and more), catching quality issues and machine downtime early, often before they happen, so less material goes to waste and the floor runs measurably leaner.
Across the Entire Data Estate
Gildan's data doesn't sit in one neat place. It lives in on-prem datacenters, high-volume transactional systems, a Snowflake warehouse, and live POS feeds from Walmart, Amazon, and Target. I work across that whole estate, stitching fragmented, fast-moving data into a single picture the business can actually trust and act on.
A Subcontinent in View
Before manufacturing, my "dataset" was an entire region. I worked with Sentinel-2 satellite imagery and gridded climate simulations spanning South Asia, modeling crop health and climate risk at a scale most data scientists never get near, to support climate-resilient agriculture planning.
From Prototype to Production
A clever demo is easy; making AI dependable across an enterprise is the hard part. I build multi-agent platforms on the Microsoft Agent Framework and MCP, generative-AI portals with real access control, and the governance that lets all of it ship safely, not just impress in a demo but run in production.
Experience
Manager, Data Science
GILDAN One of the world's leading publicly traded apparel manufacturers (NYSE: GIL).
Strategy & Technical Execution
- Developing an Enterprise Market Intelligence Agent via Snowflake Cortex for Corporate Development, integrating internal company data, market analytics, and SEC filings to benchmark operational performance against key competitors.
- Engineered an HR Organizational Development (OD) AI Agent using Snowflake Cortex, leveraging historical performance data and internal OD knowledge bases to automatically generate personalized coaching guides for managers.
- Created an Open Knowledge Format (OKF) Skill Bundle to serve as an AI Junior Data Scientist across any AI developer ecosystem (e.g., Claude Code, GitHub Copilot), automating routine tasks such as data validation code generation, security testing, and basic troubleshooting.
- Architected and deployed a Snowflake Cortex AI Agent for sewing operations that optimizes bi-weekly schedules by dynamically matching unassigned work orders to available capacity using style/color setup continuity, historical line efficiency, and custom business rules; minimizing changeover downtime, proactively notifying on work order aging, and driving significant operational gains.
- Designed and deployed a production-grade, centralized generative AI Knowledge portal on Azure, establishing RBAC to segregate multi-departmental knowledge bases (HR, IT Security, Operations, etc.).
- Architecting a Model Context Protocol (MCP) network that exposes AI capabilities as reusable, enterprise-grade microservices, enabling rapid composition of new solutions across teams.
Governance & Leadership
- Leading and mentoring a team of 5 Data Scientists to architect and deploy production-grade Data Science & AI solutions aligned with enterprise operational excellence.
- Established comprehensive Data Science governance frameworks and best practices, standardizing coding protocols and documentation for operational scalability.
- Streamlined production infrastructure by automating data and model deployment pipelines utilizing GitHub Actions and Prefect, ensuring high availability and fault tolerance.
- Championed the Invoice AI Agent, overseeing the lifecycle from conception to delivery, which slashed manual finance workloads by approximately 60%.
- Mentoring technical teams and streamlining workflows, ensuring high-quality deliverables and fostering a culture of continuous technical growth.
Data Scientist
GILDAN One of the world's leading publicly traded apparel manufacturers (NYSE: GIL).
Development & Technical Execution
- Engineered the creation of a Supply Chain AI Agent using Snowflake Cortex, accelerating strategic decision-making via natural language commands.
- Built advanced hierarchical forecasting models for spare parts tracking, directly preventing overstocking and ensuring plant operational continuity.
- Architected, deployed, and scaled an enterprise-grade end-to-end ML system predicting fabric color test results post-dyeing; minimize downstream off-shade errors and optimize real-time industrial operations.
- Developed End-to-End ML-based solution to predict fabric dyeing failures (machine downtime, scheduling conflict), allowing teams to intervene early and reduce costly material waste.
- Engineered a conversational AI agent (LlamaIndex, Azure OpenAI) that dynamically generates Python code on top of a custom data semantic layer and executes it in Snowflake runtime, enabling natural language querying over complex tabular datasets.
- Engineered ETL pipelines to ingest POS data from multiple sources (Walmart, Amazon, Target, etc.) into Snowflake, creating a single source of truth.
- Streamlined spinning machine OT data pipelines to provide live visibility of production performance in Snowflake.
Leadership & Project Management
- Led cross-functional project lifecycles from concept to production, acting as the core technical authority responsible for system design, stakeholder alignment, and ongoing operational stability.
- Served as the primary technical liaison between Data Science and executive business units, translating complex operational challenges into scalable AI roadmaps and production deployment plans.
Data Analyst
International Maize and Wheat Improvement Center (CIMMYT) Non-profit international research organization for maize and wheat science.
Development & Technical Execution
- Built an agricultural recommendation system utilizing predictive ML models to deliver targeted, region-specific fertilizer recommendations.
- Developed ML-driven anomaly detection models to identify irregularities and inefficiencies in regional fertilizer usage patterns.
- Designed automated NLP-based tools to scan online news for crop disease outbreaks, providing critical early warning signals.
- Developed a Logistics Optimization tool using Guided Local Search algorithms, achieving significant time and fuel cost savings.
- Processed large-scale Sentinel-2 satellite imagery and climate data to model crop health and support climate-resilient agriculture planning.
- Built interactive Streamlit dashboards to transform complex survey data into clear insights for research leadership.
Leadership & Project Management
- Orchestrated collaboration between remote data scientists, field agronomists, and regional analysts, creating a cohesive ecosystem for integrating diverse climate and field datasets.
- Supported and coordinated data strategy for the multi-country TAFSSA initiative, designing standardized protocols that unified agricultural data handling across South Asia.
Engineer
ECHO Center - edotco Bangladesh Co. Ltd. Telecommunications Infrastructure Services Company in Asia.
Development & Technical Execution
- Automated repetitive reporting tasks using Python-based bots, freeing up technical resources for higher-value work.
- Built power usage monitoring tools across telecom towers, identifying inefficiencies and reducing energy costs.
- Implemented automation of ETL processes using SQL and visualized insights through Power BI dashboards.
Leadership
- Managed and upskilled a team of OC engineers, establishing robust workflows for 24/7 infrastructure monitoring.
- Coordinated cross-functional troubleshooting and analytical reporting for major mobile network operators.
Selected Independent Projects & Research
Generative Analytics Agent | Natural Language to Code
Conversational agent generating SQL & Python code to analyze complex tabular data.
Field Boundary Detection using HED | Deep Learning & CV
Detect Field Boundaries using Holistically-Nested Edge Detection (HED) on airborne images.
Chemical Property Prediction with Hybrid-GICN | Graph Neural Nets
Physicochemical properties prediction with Hybrid-GICN.
Technical Skills
Languages & Big Data
ML & Modeling
Generative AI & Agents
Data Engineering & Processing
Databases, Warehouse & Cloud
Geospatial & Visualization
MLOps & DevOps
Mentorship & Supervision
Scientific Programming with Python
Tailor-Made Training (TMT+) Programme BARI, Bangladesh | 2021Geospatial Data Analysis & ML
Spatiotemporal Machine Learning with Python BARI, Bangladesh | 2021Project Supervision
Geo-Information Science and Earth Observation ITC, University of Twente, Netherlands | 2022Certifications & Training
Spatial Regression in R
GEMS Learning CIMMYT & Univ. of Minnesota | 2023Geostatistics and Interpolation in R
GEMS Learning CIMMYT & Univ. of Minnesota | 2023Open Source Scientific Computing
AgroGeospatial Big Data Analysis ITC, University of Twente | 2022Deep Learning Specialization
DeepLearning.AI 2020Algorithms for DNA Sequencing
John Hopkins University 2020Food Security and Sustainability
Crop Production Wageningen University & Research | 2020Data Visualization with Seaborn
DataCamp 2019Data Science & Big Data Foundations
IBM 2017Selected Publications
-
Advancing crop disease early warning in South Asia by complementing expert surveys with internet media scraping [Link]
Climate Resilience and Sustainability, 3, e78. (2024)
-
The optimization of conservation agriculture practices requires attention to location-specific performance: evidence from large scale gridded simulations across south asia [Link]
Field Crops Research, Elsevier (2022)
-
Climates' dual grip: Mapping upper and lower bounds for thermal stress risk for rice, wheat, and maize in Bangladesh [Link]
Technical Report (2023)
-
Personality detection from text using convolutional neural network [Link]
ICASERT, IEEE (2019)
-
Innovative new crowd sourcing tool for gathering crop disease reports developed in South Asia [Link]
ARRCC Programme Newsletter (2020)
Get In Touch
Feel free to reach out for collaborations or opportunities.