CV

My professional journey -- from Naval Architecture to Machine Learning leadership.

Contact Information

Name Shad Humydee
Professional Title Senior Data Scientist
Email showmmos@gmail.com
Phone +8801675760315
Location Remote, UK Team

Professional Summary

ML-focused Software Engineering Leader with 7+ years of experience specializing in NLP, Agentic AI, and Recommender Systems. Delivered production ML systems serving 10M+ users across enterprise and government organizations.

Experience

  • 2023 -

    Remote, UK Team

    Senior Data Scientist
    Optimizely
    Architecting production ML systems for AI evaluation, agent memory, recommendation engines, and LLM-powered platforms.
    • Built OPAL Evals — AI evaluation framework using LLM-as-a-Judge (Gemini, Claude) across 5 quality dimensions, reducing agent debugging from 2-3 days to under 4 hours
    • Architected Synapse — hybrid AI agent memory system processing 10K+ agent interactions/day with FastAPI + Python 3.13
    • Engineered CMAB recommendation engine with XGBoost handling 5K requests/second, securing major US financial services client
    • Launched AI-Segments — RAG-powered segmentation platform reducing campaign prep time by 70%
    • Built AI-Backend — multi-tenant API gateway across Azure OpenAI, Vertex AI, Replicate with GKE deployment (10 replica pods)
    • Led cross-functional teams of 3-5 engineers, increasing team efficiency by 40%
  • 2023 - 2023

    Hybrid, Dhaka

    Software Engineer II (AI/ML)
    Infolytx Inc.
    Led architecture and delivery of healthcare NLP platforms.
    • Led Avicenna Healthcare NLP platform using spaCy, SciSpacy, AWS Comprehend Medical — processing 500K+ patient records
    • Conducted R&D integrating NLP/ML with AWS for healthcare applications
    • Mentored 2-3 junior engineers, scaling team capability by 30%
  • 2022 - 2022

    Hybrid, Dhaka

    Software Engineer I (AI/ML)
    Infolytx Inc.
    Shipped healthcare ML product from demo to MVP.
    • Shipped healthcare ML product using scikit-learn, TensorFlow — reducing development-to-deployment cycle by 60%
    • Built scalable ML/NLP pipelines processing 100K+ EMR records/day
  • 2021 - 2021

    Hybrid, Dhaka

    Junior Software Engineer (AI/ML)
    Infolytx Inc.
    Built NLP solutions on EMR data for clinical insight extraction.
    • Built NLP solutions using NLTK, spaCy, NegSpacy — improving clinical insight extraction accuracy by 25%
    • Ensured HIPAA compliance through security best practices and production-ready ML workflow documentation
  • 2021 - 2021

    Remote, Dhaka

    Machine Learning Engineer
    Apurba Technologies Ltd.
    Developed OCR pipelines and predictive ML models.
    • Developed Bangla OCR preprocessing pipelines using OpenCV, Pillow, Cycle-GAN — improving accuracy by 29%
    • Built predictive ML models processing 10K+ daily data points
  • 2018 - 2021

    Dhaka, Bangladesh

    Senior Business Analyst & Business Analyst
    Integrated Software and Technologies Ltd.
    Secured high-ticket contracts and led technical documentation for government and enterprise systems.
    • Secured $2.5M project win — clients include IOM, UNDP, UNODC, WFP and government bodies
    • Led technical documentation reducing onboarding time by 35% and project delays by 20%

Education

  • 2023 -

    Bangladesh

    MSc
    United International University
    Computer Science and Engineering
    • Machine Learning
    • Deep Learning
    • Natural Language Processing
    • Data Mining
    • Computational Intelligence
  • 2014 - 2018

    Bangladesh

    BSc
    Military Institute of Science and Technology
    Naval Architecture and Marine Engineering

Awards

  • 2024
    Judge, Bangla LLM Mathematical Reasoning Challenge
    BUET CSE

    Evaluated 15+ teams building Bangla language models for mathematical reasoning.

  • 2016
    World Finalist, European Rover Challenge 2016
    European Space Foundation, Poland

    Ranked 22nd globally as technical member of Team Mongol Barota.

Skills

Agentic Workflows & Orchestration (): LangChain, RAG Architectures, LLM-as-Judge, Prompt Engineering, Tool Calling, Multi-step Reasoning
LLM APIs (): OpenAI (GPT-4o, Embeddings, DALL-E), Google Gemini, Anthropic (via Vertex AI)
Machine Learning (): XGBoost, TensorFlow/Keras, Scikit-Learn, Transformers (BERT, T5), CMAB, Recommender Systems, SHAP/LIME
Data Science & NLP (): NumPy, Pandas, spaCy, SciSpacy, NLTK, Matplotlib, Seaborn
Cloud & Infrastructure (): GCP (Vertex AI, GKE, BigQuery, Cloud Run, Pub/Sub), AWS (DynamoDB, S3, EC2), Docker, Terraform, Kubernetes
Languages & Frameworks (): Python (FastAPI), TypeScript, Java
MLOps & CI/CD (): MLflow, Weights & Biases, GitHub Actions, Bitbucket Pipelines, Jenkins, Datadog

Languages

English : Native/Bi-Lingual Proficiency
Bengali : Native/Bi-Lingual Proficiency

Interests

Artificial Intelligence: Agentic AI, NLP, Recommender Systems, LLM Evaluation, Healthcare ML

Certificates

  • Diagnosing Application Bugs with Datadog APM - Datadog (2025)
  • Veracode Security Labs Champion Level 1 - Veracode (2025)
  • Introduction to PCI DSS for Developers - Veracode (2025)
  • Software Security Awareness - OWASP Top 10 2021 - Veracode (2025)

Projects

  • OPAL Evals

    Production-grade AI evaluation framework scoring conversational agent outputs using multi-model LLM-as-a-Judge across 5 quality dimensions.

    • Reduced agent debugging from 2-3 days to under 4 hours
    • 5x faster prompt iteration cycles
  • Synapse

    Hybrid AI agent memory system for Optimizely’s agent orchestration platform, enabling personalized AI experiences through long-term memory.

    • Processing 10K+ agent interactions/day
    • FastAPI + Python 3.13
  • CMAB Recommendation Engine

    Contextual multi-armed bandit recommendation engine with XGBoost for real-time ML predictions.

    • 5K requests/second
    • 15% regret reduction
    • 8% conversion boost
  • AI-Segments

    LLM-driven segmentation platform leveraging RAG architecture and transformers.

    • 70% reduction in campaign prep time
    • 5M+ user segments/month
  • AI-Backend (Eleanor)

    Production-grade multi-tenant API gateway serving Optimizely’s AI platform across Azure OpenAI, Vertex AI, and Replicate.

    • 10 replica pods on GKE
    • Auto-scaling
    • Datadog APM observability
  • Avicenna Healthcare NLP

    Comprehensive NLP platform for EMR data analysis using spaCy and AWS Comprehend Medical.

    • 500K+ patient records processed
    • 50% reduction in physician review time