Faizul Islam

Data Scientist · Applied AI Engineer

AI that turns data into decisions, products and growth.

Smart systems. Real ROI.

I build AI-powered products end to end and plug them into how a business runs: machine learning, LLM and computer-vision pipelines, backend engineering and intelligent automation, taken from idea to production across SaaS platforms and analytics tools.

View work Get in touch
Faizul Islam

01 / About

About

Data Scientist who turns real-world data into production AI systems. I build end-to-end machine-learning and SaaS products for industry: production backends, analytics engines, computer-vision pipelines and AI-driven automation, owning the path from raw data to shipped, revenue-generating features. A first-class Statistics background gives me the rigour; full-stack engineering lets me ship it. The result is AI that earns its place in the business.

Based in
Bangladesh
Focus
LLM systems · Computer vision · Production ML · SaaS backends
Currently
Building production AI at Onwards Analytics (Australia-based)
Languages
Bengali · English · Hindi · German (basic)

02 / Experience

Experience

Onwards Analytics logo

Onwards Analytics

Australia (Remote)

Data Scientist

Nov 2025 – Present
  • Worked on a production-grade AI computer vision pipeline for real-time biomechanical pose estimation, leading full-lifecycle delivery across system architecture, model integration (MediaPipe, RTMPose, ViTPose, YOLOv8), metric engineering, benchmarking, and iterative optimisation that improved model accuracy and inference reliability.
  • Engineered and shipped the production backend of Nexlytics, a B2B data-analytics SaaS, co-building its audit-trail and scoring engines and the RAG pipelines behind its AI features.
  • Co-lead engineering and production launch of Mockingbird, a multi-tenant AI ad-intelligence SaaS: architected the brand-neutral LLM/vision generation engine (Gemini, Ideogram), enforced RLS-based tenant isolation across a Cloudflare Workers + Supabase/Postgres backend, shipped the daily-drop scoring and preference-learning loop, and hardened the codebase through a full security and repository audit ahead of launch.
  • Drove full-stack engineering and go-live delivery of MobiPet, a live two-sided veterinary marketplace (React/TypeScript, Supabase/Postgres with RLS and Edge Functions, Stripe, Cloudflare): shipped production fixes across authentication, booking integrity, geolocation and admin workflows, and built guarded data-migration and release-verification tooling to protect live customer data.

Data Scientist Intern

Aug 2025 – Oct 2025
  • Architected agentic AI and automation pipelines (n8n, Power Automate, MCP servers) that ran end-to-end consulting-delivery workflows for enterprise clients, alongside product-strategy and go-to-market documentation.
  • Delivered cross-functional engineering across Playwright-based automated UI/UX testing, backend infrastructure, GitHub Actions–scheduled web scraping and full-lifecycle SaaS QA.
DataKothon Research Lab logo

DataKothon Research Lab

Bangladesh

AI Research Intern

Jun 2025 – Aug 2025
  • Drove the full modelling lifecycle (data preparation, feature engineering, model development and evaluation) on large real-world research datasets, translating open research questions into reproducible machine-learning experiments.

03 / Work

Work

Production systems

Built at Onwards Analytics

Mockingbird

Co-lead engineer · Multi-tenant AI SaaS

What it isAn ad-intelligence platform that watches competitor Meta ads every day and turns the best ones into on-brand creative for each customer.

What I ownedThe generation engine and its tenant model, the scoring and preference-learning loop that makes each daily drop sharper, and the pre-launch security and codebase audit. Now co-leading the push to public launch.

Cloudflare WorkersSupabasePostgreSQL/RLSGeminiIdeogramTypeScriptStripe
Private · launching

MobiPet

Full-stack engineer · Two-sided marketplace

What it isA live marketplace where pet owners book after-hours home visits and mobile vets claim the jobs.

What I ownedTaking the product through UAT to go-live: auth and role routing, booking and payment-adjacent integrity, geolocation, admin tooling, and the guarded migration and release-verification scripts that keep live customer data safe.

ReactTypeScriptSupabaseEdge FunctionsStripeCloudflare
Visit site

Nexlytics

Backend engineer · B2B analytics SaaS

What it isUploads a Power BI report and returns a health score, compliance audit, issue list and documentation.

What I ownedThe production backend end to end, including the audit-trail and scoring engines and the RAG pipelines that power its AI explanations.

FastAPIPythonSupabaseRAGStripe
Visit site

AquaCoach

ML engineer · Real-time computer vision

What it isA swimming technique-analysis platform: a biomechanical pose-estimation pipeline that runs on live video and feeds downstream coaching analysis.

What I ownedFull lifecycle: architecture, model selection and integration across MediaPipe, RTMPose, ViTPose and YOLOv8, the metric layer, benchmarking, and the optimisation loop that lifted accuracy and inference reliability.

MediaPipeRTMPoseViTPoseYOLOv8Python
Private

Also built at Onwards

OnwardsOS

Internal AI-PMO platform: security audit that found and fixed 7 RLS-bypassing views, dedup logic, revived the CI migration pipeline.

ReactSupabasePLpgSQL

Consulting-delivery automation

Agentic n8n / Power Automate pipelines running end-to-end delivery workflows for enterprise clients.

n8nPower AutomateMCP

QA & scraping infrastructure

Playwright UI-test framework and GitHub Actions–scheduled scrapers across the product suite.

PlaywrightGitHub ActionsCI/CD

My Projects

Nearly

AI Visual Product Search Engine

Built an end-to-end visual search system: EfficientNet-B0 transfer learning for 1280-D embedding extraction, a FAISS vector index for top-K nearest-neighbour retrieval over fashion catalogues, and a Streamlit front end with metadata-aware recommendation re-ranking.

PyTorchEfficientNetFAISSStreamlit
Code

PricePulse

Unsupervised Anomaly Detection for Retail Prices

Architected an anomaly-detection service with PyTorch autoencoder scoring, FastAPI REST backend, Playwright scrapers for JS-rendered pages, SQLite time-series store and GitHub Actions cron automation, flagging statistically anomalous price movements across 150+ Apple products.

PyTorchFastAPIPlaywrightSQLiteGitHub Actions
Code

UrbanFlow

Analytics & ML on Ride-Sharing and Food Delivery

Engineered a Python + SQL analytics pipeline over multi-service operational data, spanning cleaning, EDA, statistical hypothesis testing and recommendation modelling, to surface demand and service-quality drivers.

PythonSQLPandasScikit-learn
Code

Querious

Advanced SQL practice log

Sustained deep dive into advanced SQL (window functions, CTEs, query optimisation) solving business problems across retail, e-commerce, finance and FMCG datasets.

SQLPostgreSQLAnalytics
Code

CO₂ Emission Forecasting (2030)

Time-series forecasting

Time-series forecasting (Prophet, ARIMA) of country-level CO₂ emissions with trend decomposition and scenario analysis.

ProphetARIMATime series
Code

More Projects

Projects on GitHub Notebooks on Kaggle

04 / Research

Research

Exploring the Impact of Malnutrition, Fever and Acute Respiratory Infections on the Physical Development of Children under Five in Bangladesh, Nepal and Pakistan (Punjab): Evidence from MICS 2019

International Conference on Applied Statistics & Data Science (ICASDS), 2025 Best Poster

05 / Skills

Skills

Languages

PythonSQLR

Machine Learning & AI

Machine LearningDeep LearningComputer VisionNLPLLMs (Gemini, Claude, GPT)RAGGenerative AIAgentic AI (Claude Agent SDK, MCP)LLM pipeline design

Libraries & Frameworks

NumPyPandasScikit-learnPyTorchTensorFlow/KerasFAISSMediaPipeMatplotlibPlotly

Backend, Cloud & Databases

FastAPIREST APIsPostgreSQLSupabase (RLS, Auth, Edge Functions)Cloudflare Workers & PagesStripeResendCloud Deployment & Migration

DevOps, Automation & Testing

GitGitHub ActionsCI/CDwranglern8nPower AutomateMCP ServersPlaywrightVitestPostman

BI, Analytics & Statistics

Power BITableauStatistical ModellingHypothesis TestingSPSSSTATASAS

06 / Education

Education

Bangladesh Navy School & College logo

Higher Secondary

Bangladesh Navy School & College, Mongla · 2018–2019

GPA 5.00 / 5.00

07 / Achievements

Achievements

  • Vice PresidentSUST Data Science Club
  • Best Poster AwardICASDS 2025, University of Dhaka
  • 2nd Runner-upData Innovators Challenge 2024
  • 2nd Runner-upML Olympiad 2024
  • FinalistRobi Datathon 3.0 (2024)
  • Kaggle 2× Expert
  • University merit scholarships
  • 1st Runner-upAstrophysics Olympiad
  • ChampionBiggan Joyotshob (science fair)
  • 5th PlaceNational Science Olympiad (Regional)

08 / Contact

Let's build something that ships.

Open to data science and applied AI roles, research collaborations, and interesting production problems.

faizulislam095@gmail.com