Mohamed Ifreen
MI

Applied Scientist · Amazon

|

About About
Mohamed Ifreen

Here's a little background

I grew up convinced the most interesting problems are the ones nobody has cracked yet. That conviction pulled me toward AI — not the hype, but the genuine possibility of building systems that reason and create real value. I completed my MS in Artificial Intelligence at Boston University with a 4.0 GPA, while simultaneously doing research and teaching ML to graduate students.

Before grad school, I spent three years at LTIMindtree as a Senior Data Engineer building NLP pipelines that cut manual workloads by 60%, and a year at Rapid Acceleration Partners as a Senior ML Engineer where my work helped secure major enterprise contracts.

I'm currently doing applied ML research at Amazon in Seattle. I'm equally comfortable reading a transformer paper on a Friday night and deploying a model on Monday morning.

Experience Experience

Applied Scientist Intern

Amazon · Creative X

Dec 2025 – Present · Seattle, WA

  • Conducting research in applied machine learning within the Creative X team.
  • Collaborating with senior scientists on developing and evaluating models at production scale.
  • Working at the intersection of NLP and large-scale intelligent systems.

Research Assistant

BU GDP Center — DATA Team

May – Dec 2025 · Boston, MA

  • Architected deep learning pipelines for the source library tool, automating policy research workflows.
  • Deployed LLM-backed agentic systems using Ollama and vLLM, significantly reducing analyst effort.

Graduate Teaching Assistant

Boston University

Jan – May 2025 · Boston, MA

  • Supported CS 542 (Machine Learning) and CS 541 (Applied ML) — two of BU's most demanding graduate courses.
  • Led lab sessions, guided students in core ML concepts, and coordinated exams.

Senior ML Engineer

Rapid Acceleration Partners

Sep 2023 – Aug 2024 · Chennai, India

  • Built a complaint-analysis AI for Hussmann (Panasonic) correlated to serial number data, enabling a major enterprise engagement.
  • Designed medical AI for VirtueServe: ICD code mapping and clinical summaries, cutting scribing time by 70%.
  • Developed voice analytics models eliminating 100+ hours of processing per cycle.

Senior Data Engineer

LTIMindtree

Oct 2020 – Aug 2023 · Bangalore, India

  • Built NLP pipelines for IT ticket mining, reducing manual effort by 60%.
  • Designed auto-tagging systems with Sentence Transformers, improving categorisation speed by 40%.
  • Received the Einstein Award and Beacon Exemplar Award for technical excellence.

← scroll for more →

Skills Skills

I've built production systems across the full ML stack — from fine-tuning large language models to shipping APIs that scale.

🐍
Python
🔥
PyTorch
🧠
TensorFlow
🤗
HuggingFace
💬
NLP / LLMs
🎯
LoRA / PEFT
🔍
RAG Systems
🤖
Agentic AI
☁️
AWS
🗄️
SQL
📊
Streamlit
🎮
Reinforcement Learning
Google TensorFlow Developer AWS Cloud Practitioner Best Paper · Boston University Hackathon Winner
Projects Projects
01
🎨
Generative AI · Bias
FLUXLoRACLIPLLaVA-NextPyTorch

Bias Mitigation in Text-to-Image Models

Fine-tuned FLUX using LoRA (rank 4) on a curated 2,700-image dataset spanning 135 professions with diverse ethnic and age representation. Built a fully automated evaluation pipeline using CLIP and LLaVA-Next across 300 test samples to verify measurable reductions in gender and skin-tone bias.

→ Boston University · November 2025

02
🛡️
LLM Security
TransformersEncoder ModelsNLP SecurityPyTorch

LLM Manipulation Detection

Explored systematic vulnerabilities in large language models and developed encoder-based detection techniques to identify and resist adversarial prompt manipulation — jailbreaks, prompt injection, identity spoofing. Awarded Best Paper at Boston University.

→ Best Paper Award · Boston University · December 2024

03
🏥
Medical AI · NLP
BERTTensorFlowMedical NLPICD Coding

VirtueServe Medical AI

Built an AI system that automated ICD code mapping and generated concise clinical summaries from physician notes. Reduced scribing time by 70% and significantly improved ICD assignment consistency across the platform.

→ 70% reduction in scribing time · Rapid Acceleration Partners

04
👁️
Vision-Language Models
VLMsVisual TokensTransformersPyTorch

Visual Token Efficiency in VLMs

Empirical study of visual token indicators across multiple VLMs, identifying structural redundancy patterns and quantifying efficiency gains pointing to concrete architectural improvements for how models allocate attention over visual inputs.

→ Directed Study · Boston University · Jan–May 2025

05
🏷️
NLP · Enterprise AI
Sentence TransformersEnsemble MLNLPTensorFlow

ITOps Auto-Tagging System

Developed an ensemble ML model to automatically classify IT tickets at scale, integrating Sentence Transformers for relevance scoring. Reduced manual effort by 60% and improved categorisation speed by 40% across LTIMindtree's IT operations pipeline.

→ 60% effort reduction · LTIMindtree · 2022

← scroll through projects →

Awards Awards
🏆
Best Paper Award
Awarded at Boston University for research on LLM manipulation detection using encoder-based adversarial defence techniques — December 2024.
Beacon Exemplar Award
Recognised at Rapid Acceleration Partners for rapid innovation and delivering AI prototypes that secured major enterprise contracts.
💡
Einstein Award
Awarded at LTIMindtree for exceptional technical excellence and innovation in machine learning systems deployed at scale.
🎓
4.0 GPA · MS Artificial Intelligence
Graduated with a perfect GPA from Boston University's MS AI programme while simultaneously working as a research and teaching assistant.
🥇
Inter-College Programming Winner
Placed first in a competitive inter-college programming challenge — strong problem-solving and algorithmic thinking under pressure.
Hackathon Winner
Built and shipped a winning AI prototype under time constraints — exactly the kind of environment that brings out my best work.
Contact Contact

Let's build something together

Get in touch.

Whether you want to discuss a role, a research problem, or an AI project — my inbox is always open.