Amir Toranjsimin, Biomedical AI Engineer

Amir Trs

Amir Toranjsimin

Full-stack healthcare AI, from the model to the interface.

  • Top Rated on Upwork
  • 9+ years experience
  • Open to contract work

Overview

Role
Biomedical AI Engineer · Full-Stack Developer
Availability
RemoteWorking with clients worldwide
Focus
Clinical AI · Signal & image processing · Product engineering

About

I build clinical AI, and I build the software a clinician actually opens. Those are usually two jobs held by two different people, and the handoff between them is where most medical AI quietly dies.

The research half is nine years of EEG, MRI and CT diagnostics. A reinforcement-learning approach to skin cancer diagnosis, published through IEEE. An autism screening framework built deliberately light, so it runs where there is no GPU budget. Quantitative EEG for bipolar classification, blockchain for medical image integrity, and a book on QEEG processing.

The engineering half is everything that happens after the paper. I took a precision health platform from an empty repository to production: a genomic engine holding just under a million classified variants, ACMG classification cross-referenced against ClinVar and GWAS Catalog, polygenic risk scores across a few hundred conditions, and role-based access for admins, doctors, technicians and patients. FastAPI and PostgreSQL underneath, Next.js on top, containerised and cloud-deployed with audit logging and HIPAA and GDPR constraints designed in from the first commit.

Selected work

(3)
Lumen health overview dashboard in dark theme

Lumen

2026

A lab report PDF becomes a dashboard you can read in ten seconds — parsed, scored, charted and interpreted, with a designed PDF back out.

  • Drop in one PDF or several. No account, no sign-up, no intake form.
  • Bubble sizes are proportional to how many markers landed in each state, and the fingerprint grid puts one dot per biomarker so an unhealthy panel is visible before you read a word.
  • Every marker is plotted on its own reference-range track, so 158 reads as how far out of range it is, not merely that it is.
  • Next.js 16
  • React 19
  • TypeScript
  • Tailwind v4
  • FastAPI
  • pdfplumber
  • jsPDF
  • LLM
Precision Health Platform dashboard showing health score, clinical metrics and genetic risk distribution

Precision Health Platform

2024 — Present

A clinical genomics platform built from an empty repository to production — a million classified variants, ACMG classification, polygenic risk scores, and a physician-configurable AI layer on top.

  • Raw genotype files classified against ACMG guidelines, cross-referenced against ClinVar and GWAS Catalog.
  • Polygenic risk scores across a few hundred conditions, alongside drug-gene interaction analysis for pharmacogenomics.
  • Multi-omic integration: genetics, clinical chemistry, microbiome, mitochondrial function, epigenetic age and wearable data in one model of the patient.
  • Next.js
  • FastAPI
  • PostgreSQL
  • Genomics
  • ACMG
  • PRS
  • Pharmacogenomics
  • Docker
  • RBAC
  • HIPAA
GoldFeed market dashboard with candlestick chart and trading signal panel

GoldFeed

2026

A Persian-language gold market terminal — live fund pricing, eleven technical indicators voting on one mechanical rating, news impact, and an analyst that draws on the chart.

  • Eleven indicators plus news impact vote into one rating — strong buy through strong sell — with an entry range, stop loss, two targets and a risk-to-reward ratio.
  • Levels come from ATR and pivots. Structure is only used when support sits closer than the ATR distance, which fixed a first version that placed stops ten percent away and dragged risk-to-reward down to 0.24.
  • If risk-to-reward falls under 1.5 the panel says so outright rather than dressing up a weak setup. Under 45% confidence, or a neutral rating, it proposes no levels at all.
  • Next.js 16
  • React 19
  • TypeScript
  • FastAPI
  • Technical Analysis
  • RTL
  • Persian
  • PWA
  • LLM

Also built

(7)

Ebo Cafe

2026

A digital café menu with an admin panel — categorised menu, live price and photo editing, specials, dark theme, fully responsive.

  • HTML
  • CSS
  • JavaScript
  • Persian
  • Responsive

ADHD EEG Detection

2024

EEG classification separating ADHD from healthy control children, using EEGLAB preprocessing and deep transfer learning.

  • MATLAB
  • EEGLAB
  • ResNet50
  • GoogLeNet
  • EEG

Hip MRI Classification

2025

End-to-end pipeline classifying hip MRI as normal or osseous lesion, with ResNet101 feature extraction and fine-tuning.

  • Python
  • ResNet101
  • Medical Imaging
  • Transfer Learning

Skin Lesion Classification

2024

Feature extraction and classification of skin lesions on the PH² dataset.

  • MATLAB
  • PH² Dataset
  • Feature Extraction

EEG Emotion Analysis

2024

EEG feature extraction, topographic mapping and affective state classification with EEGLAB.

  • MATLAB
  • EEGLAB
  • Feature Extraction

Bone Edge Detection

2024

Edge detection and contour overlay on bone imagery — Gaussian filtering, Sobel, Hough transform and active contour segmentation.

  • MATLAB
  • Hough Transform
  • Active Contour
  • X-Ray

Liver Tumor Segmentation

2024

A deep learning model that outlines liver regions in CT automatically — a task that is otherwise slow and manual.

  • Python
  • Segmentation
  • CT

Experience

Precision Health Platform

Independent project · Remote

Founding Engineer — end to end

Current

2024 — Present · Independent

  • Built the largest system I have shipped: patients register, complete intake, then upload blood work, microbiome panels and raw genotype files, and the platform runs continuous analysis rather than issuing a single report.
  • Wrote the genomic engine at its core — close to a million classified variants, ACMG-guideline classification cross-referenced against ClinVar and GWAS Catalog, polygenic risk scores across a few hundred conditions, plus drug-gene interaction analysis.
  • Built parsers for lab and microbiome reports including MITOswab and GI Effects, turning dense lab output into something a clinician can read, in English or Persian.
  • Designed the AI layer as an ensemble: several models run together where that measurably improved reliability, and a single model handles the tasks where it did not.
  • Shipped a personalised medical assistant that adapts to user feedback, plus a layer where a physician configures their own specialised agent for their patients.
  • Role-based access control and audit logging across admins, doctors, technicians and patients; interactive 2D and 3D visualisations mapping findings onto the body; exportable patient-ready clinical reports.
  • Containerised and cloud-deployed, with HIPAA and GDPR constraints designed in from the first commit rather than retrofitted.
  • Next.js
  • FastAPI
  • PostgreSQL
  • Genomics
  • ACMG
  • LLM Ensemble
  • RBAC
  • Docker
  • 3D Visualization

Freelance — Upwork

Remote · Contract

AI & Signal Processing Engineer

Current

2019 — Present · Freelance · Top Rated

  • Applied AI and signal processing for clients, most often in healthcare-adjacent domains. Top Rated with a consistent record of delivered projects.
  • Built a reinforcement learning model for optimising power generator scheduling across a 72-hour horizon.
  • Debugged a self-supervised contrastive learning pipeline for EEG motor imagery classification, then built a separate full pipeline from raw signal to trained model.
  • Found and fixed a subtle filtering bug in that pipeline that had been quietly corrupting results before anyone noticed the numbers were wrong.
  • PyTorch
  • EEG
  • Reinforcement Learning
  • Self-Supervised Learning
  • Signal Processing
  • Python

Biomedical Research

Academic research · Academic

Researcher — Medical Signal & Image Processing

2017 — 2024 · Research

  • Published across cancer diagnosis, autism detection, bipolar classification and blockchain applications in healthcare — 19 citations, h-index 2.
  • Authored a book on EEG and quantitative EEG (QEEG) processing and its clinical applications.
  • Ran diagnostic pipelines over EEG, MRI, CT, ECG and PPG data, from preprocessing and artefact removal through to trained, validated models.
  • MATLAB
  • EEGLAB
  • Deep Learning
  • Medical Imaging
  • Research

Stack

(53)
01

Language

  • Python
  • MATLAB
  • TypeScript
  • JavaScript
  • SQL
02

Machine Learning & AI

  • PyTorch
  • TensorFlow
  • Keras
  • scikit-learn
  • CNNs & LSTMs
  • Reinforcement Learning
  • Self-Supervised Learning
  • Transfer Learning
  • Hugging Face
  • Claude
  • OpenAI
03

Signal & Image Processing

  • EEG · ECG · PPG
  • EEGLAB
  • Wavelet Transform
  • OpenCV
  • Medical Imaging
  • DICOM
  • Hough Transform
04

Backend & Data

  • FastAPI
  • PostgreSQL
  • NumPy
  • pandas
  • Matplotlib
  • Docker
  • Redis
  • NGINX
05

Frontend & Product

  • React
  • Next.js
  • Tailwind CSS
  • Motion
  • D3 & Recharts
  • Three.js
  • jsPDF
  • Vercel
06

Design & Workflow

  • Figma
  • UI/UX Design
  • Design Systems
  • Motion Design
  • Data Visualization
  • Git
  • GitHub
  • Jupyter
07

Domain

  • Clinical Genomics
  • ACMG Classification
  • Pharmacogenomics
  • ClinVar & GWAS Catalog
  • Polygenic Risk Scores
  • HIPAA & GDPR

Education

Sadjad University of Technology

Master's degree · Biomedical Engineering

  • Medical Signal Processing
  • Medical Imaging
  • Deep Learning
  • Pattern Recognition
  • MATLAB

Biomedical Engineering

Bachelor's degree · Biomedical Engineering

  • Biosignals
  • Anatomy & Physiology
  • Instrumentation
  • Statistics

Continued learning

  • Python Programming
  • Statistical Analysis with SPSS
  • ReactJS Development
  • MATLAB for AI & Graphics

Publications

(4)
19
Citations
2
h-index
1
i10-index
Book · Author

EEG and Quantitative EEG (QEEG) Processing and Clinical Applications

A book on QEEG processing and how it is used clinically.

Applied research

(14)
  • Parkinson's disease detection

    Imaging

    CNN + LSTM over MRI brain scans, targeting early-stage patterns

  • Epilepsy detection system

    Signals

    Full pipeline simulated through to LCADC hardware converter integration

  • Fracture detection in X-rays

    Imaging

    Hough-transform image analysis for long bone fractures

  • Lung condition diagnosis from CT

    Imaging

    Comparative study of GoogLeNet against Inception-v3

  • EEG clustering and pattern analysis

    EEG

    Preprocessing, feature extraction and k-means, backed by PSD analysis

  • QRS detection in noisy ECG

    Signals

    Robust heartbeat detection validated on the MIT-BIH Arrhythmia Database

  • Automated hip MRI classification

    Imaging

    Preparation, augmentation and enhancement, then normal vs osseous lesion

  • ADHD brain signal classification

    EEG

    Signal processing and ML over EEG recorded during visual attention tasks

  • PPG signal analysis

    Signals

    ECG annotation techniques applied to peak detection in photoplethysmography

  • Face recognition using SVD

    Imaging

    Singular value decomposition applied to a medical imaging context

  • Liver segmentation

    Imaging

    Deep learning outlining liver regions in CT automatically

  • Heart sound classification

    Signals

    Segmentation and classification of recordings to flag abnormal patterns

  • Generator scheduling optimisation

    Systems

    Reinforcement learning over a 72-hour scheduling horizon

  • EEG motor imagery classification

    EEG

    Self-supervised contrastive pipeline debugged, then rebuilt end to end

Get in touch

Available for contract work on clinical AI, medical signal and image processing, and the product engineering around them. The messy, high-stakes end of the problem is the part I want.

Amir Trs Biomedical AI Engineer

Built with Next.js and Tailwind · 2026-08-26