Portrait of Fahim Faiyaz

FAHIM FAIYAZ

AI/ML Engineer
Upwork Fiverr Message on WhatsApp
About me

Intro

Hey, I’m Fahim. I’m a Machine Learning Engineer at Masharah Advisory, where I build pipelines that turn classical Islamic legal texts into structured digital knowledge. Before that, I spent over a year as a Computer Vision Engineer at Quantigo AI. There I validated and deployed detection models for real‑world use.

I also help businesses put AI to work. I build AI agents and chatbots that answer questions from their own documents. My other work covers automation workflows, document processing and OCR systems and computer vision models. I take each project from idea to a working solution and test its accuracy before handover. Clear milestones and proper documentation mean you always know what you’re getting.

Now
Machine Learning Engineer, Masharah Advisory
Research
6 IEEE Papers, 3 as first author
Download CVPDF · 2 pages
The stack
PythonPyTorchTensorFlowKerasScikit-learnFastAIOpenCVUltralytics YOLOPandasStatsmodels
LangChainLangFlowRAGTransformersEmbeddingsChromaDBPineconeOllamaGroqGeminiAI agentsLLM API integrationVLMs
FastAPIFlaskStreamlitGradion8nSeleniumWeb ScrapingSurya OCRTesseract OCRPaddleOCRHunyuanOCRPostgreSQLSupabase
DockerGitHub ActionsMLflowCVATONNX RuntimeTensorRTHugging Face SpacesAWSAzureTableauGit

WHAT I
BUILD

    • Chatbots that answer from your PDFs, manuals and help docs
    • Every answer shows the source it came from
    • Agents that answer customers, qualify leads and update your CRM
    • Runs on your website, in your own tools or as an API
    • Private setups on your own server when data can’t leave

    An assistant that knows your business. It answers your customers or your team from your own documents and shows the source. When the answer isn’t there, it says so.

    VIEW PROJECTS
    • n8n or Python workflows that connect the apps you already use
    • Emails, invoices and PDFs handled without copy-paste
    • Follow-ups by email, SMS or WhatsApp
    • AI steps for writing, sorting and pulling out details
    • Alerts and approval steps when a person needs to check

    Repetitive work taken off your team’s plate. I map the steps you do by hand today and automate them. Checks and alerts make sure nothing fails silently.

    VIEW PROJECTS
    • Scans, receipts and invoices turned into structured data
    • OCR for printed and handwritten text, in English and Bangla
    • Checks that flag misread numbers before they reach your books
    • Exports to Excel, QuickBooks or your own system
    • Private setups that run fully on your own machine

    Paperwork turned into data you can use. Each document is read, checked and sent where you need it. Anything uncertain is flagged for a quick review.

    VIEW PROJECTS
    • Custom models that detect, count or check things in photos and video
    • Real-time analysis of camera feeds
    • Prediction and classification models trained on your data
    • Explainable results, so you can see why the model decided
    • Delivered as an API, a web demo or inside your app

    Models built and tested on your own data. You see the accuracy numbers before launch, and get a deployed system your team can use.

    VIEW PROJECTS
Selected work

WORK

LiveDocuments & OCRMy product

Crosscheck

Receipts and invoices read by AI, then checked by plain code. Upload a batch and export clean rows. You only review what the checks flag.

91–94%of receipts read fully correct (CORD-v2 test and validation)
  • Flagged documents get one-click fixes on a review screen
  • Telegram and Gmail alerts through n8n; exports to CSV, Excel and QuickBooks
  • Ask about your documents in English or Bangla (25 of 26 test questions right)
Live GitHub 
Crosscheck home page
Built with
PythonFastAPINext.jsTypeScriptGeminiSupabasen8n
LiveWeb & mobile appsMy product

BashaBook

Shared accounts for Bangladeshi flats with a hired cook. It tracks every meal and every bazar or bill payment, then shows who owes whom at month end. A website and installable web app in English and Bangla.

  • Meals counted per person per day; tomorrow’s count locks at the cutoff, so the cook knows how many plates
  • At month end the app checks the cash count matches, then carries each balance forward
  • Push reminders for the meal lock, payments to confirm and the month-end close
Live GitHub 
BashaBook home page with a month ledger
Built with
Next.js 16React 19Tailwind 4PostgreSQL (Neon)Web pushVitestVercel
LiveAI agents & RAG

Legal Document AI

Drafts internal case memos from uploaded legal documents, grounded in what the documents say.

  • Reads scanned PDFs and images with OCR
  • Each memo is built from the uploaded documents
  • Your corrections become rules for the next memo
Live GitHub 
Legal Document AI memo screen
Built with
PythonRAGChromaDBFastAPIStreamlitSQLite
RepoAutomation

AI social media automation

A form starts the run: Llama 3.3 writes the post and a branded 1080×1080 image is composed. Nothing is published until a person approves it.

  • Copy and image in one run
  • Human approval before anything goes out
  • Publish step for Facebook and Instagram, tested on mock endpoints
GitHub 
The real n8n workflowForm trigger→Write post (Llama 3.3)→Compose brand image→Ask a person to approve→Approved?→Publish to Facebook→Instagram
Built with
Pythonn8nGroq (Llama 3.3)FastAPIPillow
Live · IEEEComputer vision & ML

Driver Distraction Detection

A YOLOv10 model that watches the driver and flags distraction as it happens. Published at IEEE ICCIT 2025.

98.1%mAP@50 across 12 behaviours
  • Spots 12 behaviours, from texting and phone calls to drinking and drowsiness
  • Works on photos and on video, frame by frame
  • Runs on ONNX Runtime, deployed with Docker through CI/CD
Live GitHub 
Driver distraction detection result
Built with
PythonYOLOv10ONNX RuntimeFlaskGradioDockerGitHub Actions
More projects

EmailCraft-AI screenshot

Agents research the lead, match portfolio items and draft a personalised email. A scoring step checks it before it goes out. Runs as an API in Docker.

CodeLive demo
LLM / Multi-Agent AI

EmailCraft-AI

Groq LlamaChromaDBFastAPIDocker
AI Fitness Coach screenshot

A router sends each question to a nutrition or training agent, and the profile is stored so advice stays personal across sessions.

CodeLive demo
LLM / AI Agents

AI Fitness Coach

LangChainGroqAstraDBStreamlit

Reads Bangla, English and mixed documents and answers questions about them. Everything runs on your own machine.

Code
RAG / OCR / Privacy AI

Local OCR & Dynamic RAG

Surya OCRBGE-M3OllamaFastAPI
Brain Tumor Detection screenshot

YOLOv10 and hybrid CNN models locate tumours in MRI scans, with Grad-CAM heatmaps that show why. Published at IEEE RAAICON and ICCIT 2024.

Code
Medical AI / Computer Vision

Brain Tumor Detection

YOLOv10TensorFlowGrad-CAM
Mango Ripeness Classification screenshot

Hybrid Xception–LSTM model that grades mango ripeness in six stages, with Grad-CAM to show what it looked at.

Code
Computer Vision / XAI

Mango Ripeness Classification

TensorFlowXceptionLSTM
Bangladeshi Landmarks Recognizer screenshot

DenseNet121 trained on a dataset scraped and cleaned by hand, deployed with Gradio on Hugging Face.

CodeLive demo
Computer Vision / Deep Learning

Bangladeshi Landmarks Recognizer

FastAIPyTorchGradio
Experience & education

TIMELINE

  1. Now
    Jul 2026 — Now

    Machine Learning EngineerMasharah Advisory

    Remote · Saudi Arabia

    I build AI that turns classical Islamic legal texts into structured and searchable knowledge for the Legal Knowledge Platform. I cover the whole ML cycle from data and experiments to evaluation and integration. The focus is accuracy and reliability at scale.

    Document AIOCRKnowledge processingEvaluation
  2. Mar 2025 — Jul 2026

    Computer Vision EngineerQuantigo AI

    Remote · Bangladesh

    I validated and deployed object detection models for client computer vision pipelines. I also built and maintained the annotated image and video datasets behind them. Tracking down annotation errors and edge cases made the models more robust.

    Object detectionImage & video annotationDataset QAModel validation
  3. 2025

    Data Science Career ProgramDokkho MasterCourse

    Cohort 7

    Selected through a competitive assessment. I completed four end‑to‑end capstone projects.

    Data scienceCapstone projects
  4. 2024 — 2025

    6 IEEE PapersICCIT and IEEE RAAICON

    3 as first author

    Research on brain tumour detection and distracted driving as well as tea leaf disease and mango ripeness.

    Medical imagingDriver safetyAgriculture
    See every paper 
  5. 2020 — 2025

    B.Sc. in Computer Science and EngineeringEast Delta University

    Chattogram, Bangladesh

    CGPA 3.60 / 4.00. Five times on the Dean’s List (GPA 3.5+) and once on the Vice Chancellor’s List (GPA 4.00).

    Thesis: Brain Tumor Detection Using YOLOv10 and Hybrid CNN Models with Explainable AI.

    ThesisComputer vision
Papers & courses

ACHIEVEMENTS

6 IEEE Papers · 3 as first author2024 — 2025
  • ICCIT 20251st author

    Deep Learning Approaches for Distracted Driving Detection: YOLOv10-S and EfficientNet-B0

    Driver safetyRead on IEEE 
  • ICCIT 20251st author

    Tea Leaf Disease Detection and Classification Using YOLOv11 and Transfer Learning with Grad-CAM

    AgricultureRead on IEEE 
  • IEEE RAAICON 20252nd author

    Hybrid Xception–LSTM Model for Explainable Mango Ripeness Classification Using Grad-CAM

    AgricultureRead on IEEE 
  • ICCIT 20253rd author

    Tea Leaf Disease Detection Using Explainable AI & the Fusion of Ensemble and Deep Learning

    AgricultureRead on IEEE 
  • ICCIT 20241st author

    A Hybrid Deep Learning Approach for Brain Tumor Detection Using XAI with GradCAM

    Medical imagingRead on IEEE 
  • IEEE RAAICON 20242nd author

    Deep Learning for Brain Tumor Detection Leveraging YOLOv10 for Precise Localization

    Medical imagingRead on IEEE 
Courses
  • Udemy2025

    Generative AI with LangChain & Hugging Face

    LLM apps, RAG pipelines and agent workflows.

  • Udemy2024

    Complete ML & NLP Bootcamp + MLOps

    Machine learning and NLP through to Docker deployment and MLOps.

Contact

LET’S TALK

Contact information

Tell me what you want to build. Messages come straight to my inbox.

Message on WhatsApp Send an email
Available for freelance work

Send a message

I’ll reply to the email you leave here.

Break it if you can. I’ll retrain it.