Current focus
Applied AI, NLP, and LLM systems
Md Mohaiminul Islam
Senior Data Scientist · Workers' Compensation Board (WCB), Alberta
Md Mohaiminul Islam — Senior Applied AI & Data Scientist
Applied AI that stands up to review.
8+ yrs
Applied ML/NLP, research to production
12
Peer-reviewed publications, 200+ citations
Human-labeled
Evaluation behind model claims
At a glance
Fit for regulated AI delivery.
Three current signals. Full evidence, architecture, and evaluation live in the deep dives.
Strongest result
Reviewable PII detection across regulated claim documents
Scope
ML point person for enterprise AI adoption: roadmap and release-gate input, mentoring, and hiring support
Employee support assistant
Retrieval quality measured against human-verified answers
A retrieval-augmented assistant for internal support questions, built inside an existing governed data platform so it inherited identity, access, and audit controls instead of standing up a parallel stack.
Agentic maintenance
Diagnoses failing pipeline runs and stages fixes for human approval
An agent harness reads a failed run, proposes a fix, and validates it on sandboxed data samples, so a person only ever approves a change that already ran clean; the same pattern was later extended to automated code review.
Claim-duration model
Gradient boosting chosen over transformers on the evidence
Benchmarked transformer approaches against a gradient-boosted classifier on accuracy, latency, and cost, took the lighter model, and paired it with drift monitoring and retraining alerts.
The WCB PII system.
The flagship work shows the pattern: reviewable AI, measurable output, and delivery inside a regulated document workflow. The next two notes add scale and research depth.
Flagship case · WCB-Alberta · 2023 - Current
Reviewable PII detection for claim documents
Designed and operationalized end-to-end PII detection across a large historical claim-document archive and a continuous flow of new documents, combining prompt-routed LLM inference, deterministic validation, post-processing, and reviewable evidence.
Problem
Claim records arrived as inconsistent PDF-extracted text, while downstream work needed dependable detection across a wide, fixed set of personal-information and entity types - a long-tail extraction problem where one missed span matters more than a good average.
Evaluation
Measured precision, recall, and F1 against labeled data, then reviewed outputs for evidence quality, overlap resolution, and traceability before downstream use.
Decisions
- Routed documents by type and text shape so extraction stayed specific to the document pattern.
- Combined LLM extraction with regex, dictionary, structural-data matching, and span repair to clean up entity evidence.
- Kept one declarative workflow contract so local and platform runs produced comparable artifacts, MLflow tracking, and release-gate evidence.
Review boundary
Weak spans, unsupported identifiers, runtime drift, or outputs without traceable evidence and run logging were treated as failure states.
System map
In: noisy PDF-extracted claim textOut: bounded entities, traceable evidence
Text intake: Load PDF-extracted claim text from approved tabular sources, covering both the historical archive and newly arriving documents.
Evidence stackSenior Data Scientist2023 - Current
ToolsPythonLLMsAzure DatabricksMLflow
Confidential-work summary. Internal document examples, prompt content, and sensitive process details stay out of scope.
02
Workers' Compensation Board (WCB), Alberta · 2023 - current
An employee-facing RAG assistant on Databricks
Built an employee-facing support assistant on the organization's governed data platform, then raised answer quality through retrieval engineering and a human-verified evaluation loop instead of prompt guesswork.
03
Blue Guardian Canada Inc. · 2023
Leading delivery of a multi-class mental-health classifier
Managed 4 direct-report data science interns delivering a 29-class mental-health text classifier, pairing transformer and LLM modeling with LLM-generated synthetic text and automated annotation.
04
Publication-grounded research · 2015 - 2023
Privacy-preserving biomedical prediction
Built privacy-preserving deep learning for drug sensitivity prediction on ~11K patient RNA-seq samples and contributed publication-backed biomedical machine learning research.
From messy documents to reviewed output.
The operating pattern behind every system below. Each proof page shows it applied end to end.
- Messy input
- Narrow AI assist
- Staff review
- Evaluated workflow
Databricks release gatesPII detectionEmployee-facing RAGForms workflow POC
Inside the systems.
Architecture, evaluation, and delivery in full: one page for the reviewable PII system, one for the wider Databricks platform work.
Experience and education.
Mohaiminul is a Senior Data Scientist in Edmonton, building reviewable LLM and NLP systems for claim-document workflows at WCB-Alberta. Earlier roles add transformer team leadership, annotation automation, and data-platform work, backed by graduate research in privacy-aware biomedical machine learning.
2023 - Current
WCB-Alberta
Senior Data Scientist
PII detection at document-corpus scale, Databricks pipelines, employee-facing RAG, agent-assisted forms, and an agentic maintenance harness for regulated claim workflows. Managed 2 interns, mentored 5 junior data scientists, supported 5 hiring processes, and presented recommendations to the CTO and director-level stakeholders.
2023
Quantolio
Senior Data Scientist
Python refactoring, Vision Transformer analysis, and a Streamlit portfolio-analytics prototype with time-series analysis and reinforcement-learning experiments for portfolio optimization.
2023
Blue Guardian Canada Inc.
Lead Data Scientist, ML/NLP
Multi-class mental-health text classification, synthetic-data strategy, annotation automation, and intern-team leadership.
2021 - 2022
Servier Canada
Data Scientist Intern, Mitacs
Built config-driven preprocessing and embedding pipelines over a large unstructured molecular dataset, removing the data-preparation bottleneck ahead of generative chemistry model work.
2015 - 2023
University of Manitoba
Research Assistant
Privacy-preserving biomedical ML, published research, and evaluation-led modeling across sensitive datasets.
Earlier work includes a data-science internship at Sightline Innovation and teaching in computer science.
Education
- PhD studies in Computer Science, University of Manitoba, 2018 - 2023 - candidacy completed; left for industry
- MSc in Computer Science, University of Manitoba
- BSc in Computer Science and Engineering, University of Chittagong
Applied AI, data platforms, and technical leadership.
What distinguishes the work. The resume carries the full technology inventory.
Applied AI systems
- Reviewable LLM workflows
- PII and entity detection
- RAG
- AI agents
- Databricks Vector Search
- Hugging Face + vLLM
- Human review loops
AI platforms
- Azure Databricks
- Model Serving
- MLflow
- Unity Catalog
- Release gates
- Model monitoring
- PySpark
- SQL
ML engineering
- Python
- PyTorch
- XGBoost
- Transformers (BERT)
- Clustering and unsupervised ML
- Reinforcement learning
Production quality
- LLM-as-judge evaluation
- Human evaluation loops
- Labeled holdout scoring
- Auditability and traceability
- Release governance
Technical direction and leverage
- AI roadmap and release-gate contribution
- Recommendations to CTO and director-level stakeholders
- Mentored 5 junior data scientists; managed 2 + 4 direct-report interns
- Supported 5 hiring and interview processes
- Governed release controls and shared model registry for team use
- One workflow contract spanning local and Databricks runtimes
Contact
If a reviewer can't check it, it doesn't ship.
Focus: reviewable LLM/NLP systems, Databricks pipelines, PII detection, and regulated document workflows in Canada.
Md Mohaiminul Islam · Edmonton, Alberta, CanadaApplied AI and NLPRegulated workflows