AI Project Lead · St. John's, NL

Mostafa Rafiur Wasib

I turn business problems into working AI and data systems, from the first requirements conversation to a tested, documented handover.

AI Project Lead at Genesis AI Garage. I combine business and systems analysis with hands-on engineering in Python, RAG, machine learning, Power BI and the cloud.

View my work Download résumé (PDF)
Portrait of Mostafa Rafiur Wasib
  • AWS Certified Machine Learning EngineerAssociate, issued May 2026
  • Master of Artificial IntelligenceMemorial University of Newfoundland
  • 10+ client AI engagementsHealth, retail, environment and B2B
  • Analysis to deliveryRequirements, build, testing and handover

What I do

From business need to working system

Business and systems analysis

Stakeholder interviews, current- and future-state process maps, requirements with acceptance criteria, user stories and UAT plans.

AI engineering

RAG and LLM applications, machine learning for forecasting and anomaly detection, Python and FastAPI services, and evaluation before release.

Data and BI

SQL, data modelling and Power BI reporting that turn operational data into KPIs people use to make decisions.

Delivery leadership

Leading client projects from discovery and feasibility through planning, testing, deployment support and handover.

Projects

Public projects you can try

Built on public St. John's and Newfoundland and Labrador data. The council minutes work was analysed, built and measured end to end:

  1. 1 · Business analysis Requirements and tests defined 7 functional and 7 non-functional requirements, 26 user stories, 30 UAT test cases
  2. 2 · AI engineering RAG app built to them Hybrid search, cited answers and an honest "not found"
  3. 3 · Evaluation 14 automated checks Retrieval, citations, refusals, speed, accessibility and privacy
Results go back into the traceability matrix, so every requirement shows its measured result.

AI engineering · RAG

St. John's Council Minutes Assistant

Ask about St. John's council decisions and get a short answer from a local open model that cites the meeting and agenda item, or an honest "not found". Search runs live in your browser.

Correct source in top 5
95%
Unanswerable questions declined
20 of 20
Cost per answer
$0 (local model)
  • Python
  • Hybrid search
  • Local LLM (MLX)
  • pgvector
  • FastAPI
  • Evaluation
Live demo Architecture Code

Business analysis

Council minutes search: business analysis case study

The analysis behind the app above: requirements, current- and future-state BPMN maps, user stories, a UAT plan, and traceability from every requirement to a test.

Solution requirements
52
User stories
26
UAT test cases
30
  • BABOK
  • BPMN 2.0
  • User stories
  • UAT
  • Traceability
2-page summary (PDF) Full case study

Data and BI · Power BI

St. John's economy and housing dashboard

Power BI report on Statistics Canada data comparing prices, rent, jobs, housing and population in St. John's with N.L. and Canada, plus a web version.

Statistics Canada series
61
DAX measures
53
Monthly data since
2016
  • Power BI
  • DAX
  • Power Query
  • Star schema
  • Python ETL
  • D3
Interactive version Power BI files

AI-assisted engineering · Workshop

Vibe Coding, Engineered (MVPFlow AI)

A spec-driven, AI-assisted MVP workflow and the two-hour workshop that teaches it: founder notes become requirements, architecture, a backlog and reviewed pull requests.

Automated tests
41
Workshop
2 hours
API key needed
No (offline mode)
  • Python
  • FastAPI
  • Spec-driven development
  • CI
  • Teaching
Code and workshop

Client work

Production AI work at Genesis AI Garage

My client work is under NDA, so these are brief, approved summaries of selected projects: the domain and my role, without client data, results or code. I'm glad to talk through my approach in interviews, within those limits.

StarLuv · Retail product discovery

Personalized product recommendations

Built the personalization and ranking of a retrieve, rank and explain recommendation workflow, with evaluation checks and automated tests.

  • Python
  • Vector search
  • LLM APIs
  • Evaluation

Artinus · B2B AI and data analytics

Document intelligence and RAG platform

Built and owned a cloud platform that extracts structured data from business documents, validates it and answers questions with citations, with CI quality and security checks.

  • Python
  • Azure
  • OCR
  • pgvector
  • CI/CD

OmaScan · Accessibility and 3D home assessment

3D scan intelligence

Built an engine that turns 3D home scans into structured fixture data for accessibility assessments, designed to flag uncertain results instead of guessing.

  • Python
  • 3D and vision
  • Structured outputs

DuaTask · Business operations

Cloud deployment and production readiness

Primary contributor on a document-processing platform: private cloud deployment on Azure, infrastructure as code, automated tests and handover documentation.

  • Azure
  • Python
  • Infrastructure as code
  • Documentation

Weeva · Business decision support

Evidence-verification pipeline

Building an MVP that gathers public evidence for a business question, checks every quote against its source and summarizes the findings into a recommendation.

  • Python
  • LLM verification
  • Web search
  • Evaluation

Nucliq · Healthcare and biotech

Knowledge assistant (RAG)

Built the first serverless RAG prototype on AWS and supported the production build: document ingestion, vector retrieval and guardrails for health-related answers.

  • AWS
  • RAG
  • pgvector
  • Terraform
Show 3 more client projects

Audyse · Audio AI

Emergency sound detection

Prototyped emergency-sound detection with direction estimation for noisy environments, and assessed DSP and AI trade-offs under real-time limits.

  • Python
  • Audio ML
  • ONNX

Nditive · Environmental sensor analytics

Forecasting and anomaly detection

Prototyped forecasting and anomaly-detection models for environmental sensor data and reported results through dashboards.

  • Python
  • XGBoost
  • Anomaly detection
  • Grafana

Soma Health · Clinical workflow support

Healthcare AI architecture review

Reviewed and contributed to a healthcare AI platform's architecture: retrieval, logging, deployment and cost considerations for a privacy-sensitive workflow.

  • AWS
  • RAG
  • Architecture review

Experience

From intern to AI Project Lead

  1. Sep 2024 to present

    Genesis Group Inc. / Genesis AI Garage

    St. John's, NL

    AI Project Lead Apr 2026 to present

    • Lead concurrent AI and software engagements across 10+ confidential client projects, end to end: discovery, feasibility, scoping, planning, delivery, testing and handover.
    • Run requirements and scoping sessions, turn them into prioritized backlogs, manage scope, risk and timelines, and coordinate engineers, reviews and client demos.
    • Stay hands-on: review solutions, validate outputs and troubleshoot before release.

    AI Engineer Jan 2025 to Mar 2026

    • Built AI, machine-learning, data and API solutions for clients in Python, TypeScript, FastAPI, SQL and Docker on AWS and Azure: RAG and document intelligence, forecasting and anomaly detection, and recommendations.
    • Worked directly with clients on requirements, demonstrations and issue resolution; wrote test plans, deployment notes and handover guides.

    AI Project Intern Sep 2024 to Dec 2024

    Prepared data, ran machine-learning experiments in anomaly detection and time series, and reported results using AWS analytics tools.

  2. Earlier roles

    • Data Analyst Intern, Stella's Circle and Choices for Youth (2024): KPI dashboards and integration-ready data.
    • Community Researcher Intern, Canadian Council on Rehabilitation and Work (2024): employment-access analysis.
    • Research Assistant, Visual and Analytic Computing Lab, Memorial University (2023): image and medical-data experiments.
    • IT Administrator and Data Analyst, Vision One Consultancy, Phnom Penh and remote (2022 to 2023): IT support, access management and CRM, plus dashboards, ETL and data quality.
    • Affiliate and Digital Marketer, Vision One Consultancy, remote (2020 to 2021): campaign, traffic, engagement and conversion analysis.
    • Data Science Intern, Maxis Berhad, Kuala Lumpur (2019 to 2020): outlier analysis and Power BI and Tableau dashboards.

Skills

Tools and methods I use

AI and machine learning
Python, LLM applications, RAG, LangGraph, LangChain, structured outputs, LLM evaluation, embeddings and vector search, scikit-learn, XGBoost, PyTorch, TensorFlow, forecasting, anomaly detection, recommender systems, Claude, OpenAI, AWS Bedrock
Business and systems analysis
Requirements elicitation, business requirements documents, BPMN process mapping, current- and future-state analysis, user stories and acceptance criteria, UAT planning, requirements traceability, stakeholder communication, technical documentation
Data and BI
SQL, PostgreSQL, pgvector, MySQL, Power BI (DAX, Power Query), Tableau, Excel, data modelling, data quality, ETL pipelines, KPI dashboards, campaign analytics
Cloud, engineering and delivery
FastAPI, REST APIs, TypeScript, Node.js, Next.js, React, Supabase, Docker, GitHub Actions, AWS (CDK, S3, Athena, Lambda), Azure, Google Cloud, Terraform, Grafana, Agile delivery, scope and risk management

Education

Degrees and certifications

University

Certifications

Earlier courses

Coursera, 2020: Getting Started with AWS Machine Learning (ID UVRGD9SJTMF2) · Introduction to Data Analytics for Business (ID ZWM3N6TKGKR9) · AWS Fundamentals: Going Cloud-Native (ID FLA8XC8UE6H5) · AI For Everyone (ID BNHHR2YN693L)

More projects

Academic and independent projects

Earlier work from my Master of AI and independent study: computer vision, NLP, classical machine learning and reinforcement learning.

Show academic projects (9)

  • Multimodal emotion recognition

    Facial-expression CNNs combined with Wav2Vec2 voice features to recognize emotion from video and audio.

    DetailsCode

  • DNA sequence classification

    Two-person project comparing k-mer, binary and TF-IDF encodings for gradient-boosted DNA classification; I built the k-mer pipeline, which performed best.

    DetailsCode

  • Fungal habitat prediction (Newfoundland and Labrador data)

    6,162 preserved fungal specimen records: data-quality analysis and Random Forest and LightGBM models benchmarked against a majority-class baseline.

    DetailsCode

  • Gas sensor classification and anomaly detection

    Gas-type classification and anomaly detection on sensor-array data, comparing Keras and PyTorch implementations.

    Code

  • Sentiment analysis

    Three-class sentiment classifier comparing Naive Bayes, logistic regression and SVM on TF-IDF features.

    DetailsCode

  • Extractive text summarization

    Sentence scoring by frequency, position and keywords, with sentence embeddings to avoid redundant picks.

    DetailsCode

  • Reinforcement learning experiments

    SARSA and Q-learning in grid worlds, Monte Carlo and TD(0) value estimation, and multi-armed bandit algorithms.

    Grid worldBanditsPoliciesCode

  • Real-time face, age, gender and emotion detection

    OpenCV DNN face detection combined with pre-trained age, gender and emotion models on live video.

    DetailsCode

  • Visual-to-audio accessibility extension (team project)

    Browser extension prototype that describes images aloud for visually impaired users, using image captioning and text-to-speech.

    DetailsCode

Contact

Let's talk

I'm open to AI engineering, business systems analysis and data roles in St. John's, and to hybrid or remote roles across Canada. Email is the fastest way to reach me.

Email me View résumé