Hello! I'm Muhammad Enrizky Brillian, a results‑driven machine learning and data science professional graduating May 2026, with deep expertise in building and deploying scalable AI solutions. I have a strong foundation in machine learning, data engineering, and cloud technologies, complemented by hands-on experience in developing innovative AI applications. My passion lies in leveraging data to solve complex problems and drive business value.
I specialize in building end-to-end AI solutions with expertise in NumPy, Pandas, Scikit‑Learn, PyTorch, TensorFlow, Keras, LangChain, Hugging Face, PySpark, AWS, GCP, SQL and React/TypeScript. As a Machine Learning Developer at IBM, I've architected multi‑agent Text‑to‑SQL pipelines achieving 93% accuracy and delivered multimodal RAG systems with 97% relevancy, securing over $300K in contracts.
- 🌐 Portfolio Website: https://billy-enrizky.github.io/portfolio/
- 💼 LinkedIn: Muhammad Enrizky Brillian
- 🐱 GitHub: billy-enrizky
- 📧 Email: [email protected]
- 📱 Phone: +14167317583
- Architected end‑to‑end, multi‑agent Text‑to‑SQL solution within AWS, combining Bedrock (LLM), RDS (database), Redis (Cache), achieving 93% accuracy for over $300K contract
- Developed multimodal agentic RAG solution by unifying image and text databases, agentic PDF parsing, and reranking, delivering 97% answer relevancy, invoked $100K Contract
- Enhanced model accuracy from 56% to 83% through ML model selection and hyperparameter tuning
- Achieved 91% accuracy on multimodal model, combining text and image features using early fusion
- Performed data preprocessing for 122GB of pathology medical images on Linux-based supercomputers
- Built ETL pipeline using Python and SQL, integrating billions of rows from the largest vaccine production plant in Canada ($800M worth)
- Engineered two-click, self-service workflow on Snowflake and Streamlit, reducing manual cycles from days to seconds
- Led cross-functional team, enhancing productivity by 40% through agile methodologies
- Developed full-stack platform using NodeJS, ExpressJS, NextJS, React, TypeScript targeting $12.5M annual savings
- Built paper-screening agent integrating various LLMs achieving 92% accuracy
- Implemented scalable ML pipeline using OpenCV, Pandas, MediaPipe to process 60 high‑resolution videos (7 GB each)
- Engineered multimodal AI data extraction pipeline leveraging GPT-4 Vision achieving 99.98% accuracy
- Subjects: Data Science, Calculus, and Statistics (495+ hours of experience)
- Conducted tutorial sessions for 30+ students, graded assessments for 650+ students
- Frameworks: PyTorch, TensorFlow, Keras, Scikit-Learn, Hugging Face
- Specialties: Deep Learning (94%), Computer Vision (97%), NLP, RAG Systems
- AI Agents: CrewAI, LangChain, Multi-agent Systems
- Languages: Python, R, SQL, JavaScript, TypeScript
- Libraries: NumPy, Pandas, OpenCV, MediaPipe, tidyverse
- ETL & Processing: PySpark, Snowflake, data preprocessing (96%)
- Visualization: Tableau, Matplotlib, Seaborn (93%)
- Platforms: AWS (Bedrock, RDS), GCP, Linux-based supercomputers
- Databases: Redis, SQLite, SQL databases
- Web Development: React, NextJS, NodeJS, ExpressJS, Flask, Streamlit
Developed multi-agent AI system leveraging CrewAI to analyze job postings and tailor resumes dynamically, integrating SerperDev and MDXSearch for personalized resume creation.
Engineered advanced RAG solution using LangChain with Chroma Vector Database and HuggingFace Embeddings, integrating multiple LLM providers (OpenAI, Anthropic, Groq).
Built multimodal LLM using IBM WatsonX API and Streamlit, integrating llama 3.2 90b vision instruct model for context-aware responses to visual and textual inputs.
Pioneered AI Agent Assistant leveraging CrewAI Framework and Exa API, integrating OpenAI, Groq, and Ollama models for robust research insights.
Implemented Deep Q-Network (DQN) using TensorFlow with dueling network architecture and epsilon-greedy policy for Atari games.
Implemented DCGAN with PyTorch to generate realistic faces, training on 21,551 face images with 2.7M discriminator and 3.8M generator parameters.
Developed hybrid CNN-LSTM architecture achieving 91% accuracy, implementing LRCN architecture for spatial and temporal feature integration.
Created comprehensive financial dashboard showcasing 3 years of performance with interactive visualizations of P&L statements and profit margins.
Implemented bidirectional LSTM with attention mechanisms achieving BLEU score of 0.54, utilizing VGG16 for image feature extraction.
Trained LSTM neural network with 77,160 parameters achieving 98.61% accuracy, engineered MFCC preprocessing pipeline with Streamlit interface.
Bachelor in Data Science and Machine Learning Specialist
University of Toronto (2022 - 2025)
- Funding: Advanced Indonesian Scholarship (BIM) from Indonesia Ministry of Education (~$380,000)
- Activities: Teaching Assistant, Finance & Data Lab Assistant, Academic Representative
- 📊 Data Analysis: Unlock insights and drive decisions with comprehensive data analysis
- ⚙️ Data Engineering: Build robust ETL pipelines and data infrastructure
- 📈 Data Visualization: Transform data into compelling stories with interactive dashboards
- 🤖 Machine Learning: Develop predictive models and AI solutions
- 💼 Business Analysis: Optimize operations through data-driven insights
- 👨🏫 Teaching & Mentoring: Academic support in Data Science, Statistics, and Calculus