About Me

I am a Senior Applied AI/ML Scientist with 6+ years of experience delivering production AI products across finance, telecom, and music. My recent work includes building multi-agent document intelligence, hybrid visual-lingual signature authentication, and scalable music foundation models with extended context.

I turn research into products by combining deep learning, graph-based reasoning, and large-language-model engineering. I enjoy building robust MLOps pipelines, deploying inference-ready systems, and helping teams adopt modern AI tooling.

I hold a PhD in Computer Science from the University of Warwick and collaborate across industry and research to solve hard AI challenges with practical impact.

Prior to Warwick, I studied MSC and BSC at the University of Tehran in the ECE Department, Iran. In my master's program, I worked on using deep learning models to predict stock prices. Throughout my master’s, I implemented and designed many deep learning models for different tasks.

I received my diploma in Physics and Mathematics from Shahid Beheshti, under the supervision of NODET (National Organization for Developing Exceptional Talents).

I was born in a beautiful city, Shahrekord in Iran (you can see some photos here). I’ve spent 25 years of my life in Iran, which has imparted great skills and memories.

View My Resume (PDF)

Work Experience

Senior Applied AI/ML Scientist

JPMorgan Chase, London, UK — 2024–Present

Delivered document intelligence and multi-agent systems for custody operations, with LangChain, LangGraph, and OpenAI Function Calling to orchestrate 6+ systems and maintain long-term memory.
Built a hybrid visual-lingual signature authentication pipeline that extracts mandate restrictions, detects relevant signatures, and verifies them against approved signature lists.
Deployed production-grade APIs with FastAPI, Docker, Azure OpenAI, ONNX, and automated observability using Comet and cloud monitoring.

Senior Machine Learning Engineer, Co-Founder

Emergesound.ai, Remote — 2024–Present

Training one of the first music foundation models using 130M songs and 4.7B tokens with balanced genre and style coverage.
Implemented a scalable cloud-optimized data pipeline and streaming architecture to remove training bottlenecks and support inference-ready deployment.
Extended model context from 9k to 36k tokens using RoPE positional encoding and prepared architectures for Triton, SageMaker, and Bedrock deployment.

Senior Data Scientist

Nokia, Reading, UK — 2022–2024

Delivered computer vision products for telecom tower inspection, worker hazard detection, and edge-device model conversion with quantization.
Built a face verification check-in solution and contributed to Nokia GenAI strategy as part of a company taskforce.
Optimized large-scale video training with gradient checkpointing, flash attention, and video-language supervision.

Machine Learning Engineer

DeepMirror, Cambridge, UK — 2021–2022

Applied graph solutions to low-data life-science problems across images, molecules, RNA/DNA, and antibodies.
Generated drug and protein candidates and improved segmentation with very few labeled examples.
Represented RNA data as graphs for the first time and helped secure a UKRI Smart grant.

Research Collaborator

Google AI, Collaboration — 2021

Designed novel self-supervised tasks and extended graph learning ideas across heterogeneous multi-modal data.

Machine Learning Intern

Intel AI Lab, San Diego, CA — 2020

Designed a Learnable Graph Inception Network for multimodal emotion recognition using video, audio, and sensor data.

Multi-agent Document Intelligence

AI system that routes documents through agents to extract structured facts, summarize clauses, and automate custody workflows.

Music Foundation Model

One of the first large-scale music foundation models trained on 130M songs with extended contextual representation.

Edge AI Model Conversion

Scalable edge deployment pipeline for optimized model conversion and quantized inference.

Emotion Recognition

Model-agnostic graph-based solution for different modalities

Semi-supervised Audio Learning

Self-supervised graph representation learning

Emotion Recognition

Research about emotion detection based on GCN

3D Foot Scanner

Commercial product

E-passport Reader

Commercial product

Persian Poetry Generator

Poet generating with GAN

Time Series Prediction

Time series prediction based on spectral image

Reservoir Level Control

Control system

Machine learning for epilepsy detection

Read the paper here

Epilepsy is a common neurological disorder, charactererized by abnormal firing of neurons. Magnetic Resonance Imaging (MRI) techniques can be integrated with machine learning methods to diagnose epileptic patients noninvasively. In this study, we use structural MRI data of 17 subjects (10 epileptic patients and 7 normal control subjects) and segment brain tissues using a Gram-Schmidt orthogonalization method and a unified tissue segmentation approach. We then compute first-order statistical and volumetric gray-level cooccurrence matrix (GLCM) texture features and train SVM classifiers for epilepsy diagnosis based on the features of the whole brain or those of the hippocampus. We achieve an accuracy of 94% using the unified segmentation method and whole-brain analysis approach.

Education

PhD. in Computer Science

Department of Computer Science, University of Warwick, Coventry, UK, graduated on Nov, 2022.

Highlights:
- Working on graph-based neural networks
- Reviewer at Winter Conference on Applications of Computer Vision 2023
- Reviewer at IEEE Transactions on Affective Computing journal
- Part of the Doctoral Consortium Program Committee at ICMI 2022
- Elected as a reviewer at the first dedicated graph ML conference LOG
- ICME reviewer in 2022, 2021, and 2020
- IEEE Signal Processing Letters reviewer in 2022
- Science Publishing Group, Machine Learning Research reviewer in 2022
Supervisor: Prof. Tanaya Guha

M.Sc. in Electrical Engineering (GPA: 3.54/4)

School of Electrical and Computer Engineering, University of Tehran, Tehran, Iran, 2018.

Highlight: Worked on deep learning
Related coursework: Machine Learning, Pattern Recognition, Neural Network & Deep Learining, Game Theory
Supervisors: Prof. Ahmad Kalhor and Prof. Babak N Araabi
Thesis: ”Prediction in Temporal Systems Based on Online Spectral Image Generating and Using Deep Learning Neural Networks”, you can download article version of my thesis here.

B.Sc. in Electrical Engineering (GPA: 3.48/4)

School of Electrical and Computer Engineering, University of Tehran, Tehran, Iran, 2016.

Highlight: Applied smart control method to process of anesthesia
Related coursework: Algorithms, C/C++ Programming, Introduction to Artificial Intelligence, Linear Algebra
Supervisor: Prof. Fariba Bahrami
Thesis: ”Simultaneous control of anesthesia and analgesia In the process of anesthesia during surgery ”

Research Interests

Machine Learning

Graph Machine Learning

Multi-modal Learning

Affect Computing

Teaching Experience

University of Warwick

CS345: Sensor Networks and Mobile Data Communications
CS342: Machine Learning
CS355: Digital Forensics
CS416: Optimisation Methods

University of Tehran

Neural Network & Deep Learning
Pattern Recognition
Filter and Circuit Synthesis
Engineering Mathematics
Microprocessors

Technical Skills

Programming Languages

Python (6+ years), C++, MATLAB

Machine Learning and Deep Learning Libraries

Scikit-learn, Pytorch, Tensorflow

Numerical Analysis and Optimization Libraries

NumPy, SciPy, hyperopt

Cloud Computing Platforms

AWS (EC2, S3, Lambda), Microsoft Azure (ML)

Software Development Tools

Git, Docker

Honors and awards

Smart grant of 2022 approved by UK Research and Innovation (UKRI).
Warwick Computer Science PhD students scholarship.
Ranked as top 0.15 % (442th) among around 270,000 participants in Iran’s National University Entrance exam in 2011.
Ranked as top 10 % in Cumulative GPA among all students of Electrical Engineering in 2014.
Accepted in the first stage nationwide competition to select national Mathematics Olympiad team.

Personal Interests