Experience & Academic Path

A focused journey through applied AI research and production-grade machine learning engineering.

work

Professional Experience

smart_toy

Applied AI Scientist

Homebase
Aug 2025 — Present
  • arrow_rightBuilt a production sales forecasting model achieving 18% average sMAPE and 25% average MAPE across live deployments, directly impacting scheduling decisions for thousands of small businesses.
  • arrow_rightDesigned and shipped Signal Scout — a RAG-powered feedback intelligence system that ingests NPS scores, App Store reviews, and internal survey data, maps them to product domains (scheduling, time tracking, etc.) with 80% classification accuracy, and auto-generates a weekly CEO-level insight report — boosting the company's customer feedback capture by 10x.
  • arrow_rightDeveloped an internal Streamlit platform enabling teams to explore raw customer feedback, apply filters, chat with AI for recommendations, and export PDF reports.
  • arrow_rightCurrently building guardrail infrastructure for production LLMs — starting with audio models (silence detection, language detection, noise reduction, confidence scoring), with image and text guardrails on the roadmap.
PythonRAGLLMsStreamlitForecastingNLP
shield

ML Engineer Intern

eSentire
Feb 2025 — Jul 2025
  • arrow_rightBuilt a scanning tool for Model Context Protocol (MCP) vulnerabilities, identifying 10+ critical security risks and generating detailed mitigation reports for engineering teams.
  • arrow_rightDeployed and configured 5 distinct LLM protection systems; integrated Promptfoo for red-teaming with 5,000+ adversarial prompts and implemented NOVA rules to analyze over 1M prompt logs, improving harmful prompt detection accuracy by 20%.
  • arrow_rightDevised a novel framework for jailbreak detection in LLMs by researching adversarial prompt techniques and evaluating detection methodologies across multiple threat models.
  • arrow_rightFirst-authored MCP-Scanner — the first comprehensive security vulnerability scanner for Model Context Protocol servers — accepted at EnCyCriS @ ICSE 2026 (Rio de Janeiro, Brazil).
PythonLLMsPromptfooRed TeamingMCPSecurityJailbreak Detection
school

Academic Background

menu_book

Master of Applied Science (MASc) in Computer Science

University of Guelph

Specialization in NLP & Recommender Systems

workspace_premium

Bachelor of Science in Computer Engineering

Ferdowsi University of Mashhad

Specialization in NLP & Recommender Systems

data_object

Skills & Technologies

Research Domains

NLP & Large Language ModelsRecommender SystemsAI Safety & GuardrailsForecasting & Time SeriesSocial ComputingRetrieval-Augmented Generation

Engineering Stack

PythonPyTorchFastAPIStreamlitDockerPostgreSQLDatabricksGit

ML & AI Tools

Hugging FaceTransformersLangChainscikit-learnPandasNumpySpeechBrainOpenAI API