Solutions Architect Machine Learning
Resume:
Pham Minh Long
+84-036******* # ****************@*****.*** ï LinkedIn § Syun1208 # Leon
Objective
My short-term objectives are to quickly adapt to requirements, identify and solve problems within a month, and efficiently update new technologies. In the long term, I aspire to become an AI Solutions Architect. Education
HCMC University of Technology and Education Aug 2019 – Sep 2023 Bachelor of Engineering, Computer Engineering – Graduated Valedictorian GPA: 3.25/4.00 (8.21/10.00) Top 1/60 Technical Skills
Programming Languages: Python, Rust, Bash Shell, C/C++, Java. Programming Skills: S.O.L.I.D Principle, Design Pattern, Distributed Training, Multi-threading Programming. Frameworks:
• Deep Learning: Tensorflow, Keras, Pytorch, OpenCV, ONNX, DeepSteam-YOLO, TFLite, TensorRT, WanDB, Tensorboard.
• Machine Learning / Data Mining: Pandas, Polars, Apache Spark, Scikit-learn, Seaborn, Matplotlib, Numpy.
• Application: Flask, FastAPI, Dijango, Streamlit.
• LLM: LangChain, LlamaIndex, CrewAI
• Unit Test: pytest, unittest.
Vector Database: Faiss, ElasticSearch, qDrant, Chroma Relational Database: MySQL, MSSQL
NoSQL Database: Redis, MongoDB
Release Technologies Usage:
• Workflow: Apache Airflow, Mlflow.
• LLMOps: vLLM, Ollama, Docker, Kubernetes, HAProxy, Portainer, Grafana, Prometheus.
• CI/CD Script: Gitlab-CI, Jenskin.
• Cloud: AWS, Google Cloud Platform.
• API Testing: Postman.
• System Design: Load balancing, Caching.
• Tools: Git Extension, Portainer, Grafana.
Knowledges:
• Have sturdy comprehension types of RAG, Multi-Agent System (MAS), Chunking, Model Embedding and LLM tuning techniques including LoRA, PPO, DPO, RLHF, GRPO, Multimodal Learning, EDA/Data Mining, ML/MLOps, System Design (Database Optimization, Load Balancing, Caching, API Design), Domain Adaption, Knowledge Distillation, Quantization, GNN.
• Strong ability to research scientific papers and implement the latest trend in AI technologies.
• Familiar with basic AWS, DBMS (Indexing, Schema, SP), DevOps concepts, Git workflow, CI/CD, Service Scheduling. Soft Skills: Critical-thinking, Problem-solving, Presentation, Teamwork. Work Experience
Nexcel Solutions Sep 2023 – Now
ML Engineer
• Design and implement a multi-agent system that recognizes entities from a user’s query and extracts necessary informations for report generation for software team, using Ollama and Huggingface for entity recognition tuning.
• Research, design and implement an Agentic RAG system and Web Interface for Domain-Specific Conversational AI for Betting Knowledge.
• Research and develop a data and model monitoring system focused on tracking data shifts to minimize model performance (pre-active customer classification) decay over time.
• Analyze and research building models for Customer’s Time-Series Performance Prediction.
• Apply S.O.L.I.D principles and design patterns to refactor code effectively, write unit tests, and reduce size of Dockerfiles using multi-stage build, caching. Develop stored procedures (SPs) in preparation for production performance monitoring. Implement workflows in Apache Airflow to monitor logs and streamline the release process. Set up Ollama and vLLM for serving local LLMs to ensure data security.
• Serve as a pioneer in designing and building the CI/CD pipeline, optimizing Dockerfiles with multi-stage builds, and implementing an infrastructure for service scheduling using Apache Airflow. Act as the sole team member responsible for managing Portainer and accessing VM accounts in production, monitoring, and resolving issues for all services of the AI team.
Unicloud Group Apr 2022 – Dec 2022
AI Engineer
• Design and improve Unicloud eKYC system - ID Card Verification. Wisdom Engineering and Business Solution Jan 2022 – Apr 2022 AI R&D Intern
• Design models for Speech Recognition.
UTE-AI Lab May 2021 – Dec 2023
Research Assistant
• Report, participate in countless competitions in HCMC.
• Catch on the state-of-the-art tendencies in AI fields instructed by PhD. Tran Vu Hoang. HCMUTE Jun 2020 – Aug 2020
Teaching Assistant – C Programming Language
• Supervised by PhD. Pham Ngoc Son, Department of Computer Engineering - Communication, Faculty of Electrical and Electronics Engineering.
Projects
Report Entity Recognition HuggingFace Feb 2024 – Now My responsibilities
• Research, design, and implement a multi-agent system to support multi-report generation across various departments.
• Generate and clean datasets for training and testing using LLMs.
• Write API documentation using Swagger and implement API Authentication.
• Integrate the system with Telegram to provide a user-friendly interface. Technology Description
• Pretrained Models: Qwen2.5-Instruct-7B
• Training Technique: Bidirection Transformer, Entity and Span Representation
• LLM Engine: Ollama
• Framworks: HuggingFace, FastAPI
• System Design: Load balancing (HAProxy), Caching (Redis) Team Size: 1
Company: Nexcel Solutions
My position: Junior ML Engineer
Domain-Specific Conversational Chatbot for Betting Knowledge Dec 2024 – Now My responsibilities
• Research and implement an Agentic RAG baseline and Web Interface within 1 week to power a chatbot that provides support with company-specific information and betting knowledge
• Collect, chunk, and organize metadata from PDF and HTML templates, then fine-tune Llama-3.2-7B and Qwen2.5-Instruct-7B using techniques like LoRA and PEFT.
• Build Dockerfile preparing for CI/CD pipeline, implement load balancing to manage overloading during multiple concurrent requests in the release.
• Evaluate RAG System using BERT Precision/Recall/F1-Score, Diversity, BLUE Score Technology Description
• Pretrained Models: Llama-3.2-7B, Qwen2.5-Instruct-7B
• LLM Engine: Ollama, vLLM
• Database Engine: MySQL, Faiss, MongoDB
• Framworks: Langchain, FastAPI Next.js
• System Design: Load balancing (HAProxy), Caching (Redis) Team Size: 1
Company: Nexcel Solutions
My position: Junior ML Engineer
Customer’s Time-Series based Performance Prediction Jun 2024 – Dec 2024 My responsibilities
• Comprehend the business requirements, define the research directions.
• Explore, analyze and conduct statistics on data.
• Leverage machine learning models, apply business knowledge and seek ideal features to build RFM model to support two key strategies: Customer-based and Product-based segmentation. Use these techniques to predict daily customer performance.
Technology Description
• Unsupervised Learning: K-Means, DBSCAN
• Supervised Learning: Random Forest, ExtraTrees, XGBoost
• Training Technique: Unsupervised Domain Adaption, Active Learning
• Data Mining Techniques: Rule-based method, aPriori, RFM Model
• Dimensionality Reduction: PCA, t-SNE, UMAP
• Feature Engineering: Target Encoding, Entropy, Bayes Statistics Team Size: 1
Company: Nexcel Solutions
My position: Junior ML Engineer
Data and Model Monitoring Nov 2023 – Jun 2024
My responsibilities
• Research metrics and conduct some hypothesis tests to measure distribution shifts in both reference and production data for the PAC project, one of the most critical initiatives in Nexcel Solutions’ Group in Taiwan.
• Apply S.O.L.I.D principles and design patterns to refactor code, develop unit tests, optimize the Dockerfile, and write unit test and stored procedures (SPs) in preparation for the CI/CD pipeline.
• Design an Apache Airflow pipeline to trigger tasks and monitor its daily performance to ensure smooth operation. Technology Description
• Metrics: PSI, Hellinger, Wasserstein, Energy distance, KL Divergence, JensenShannon.
• Hypothesis Tests: KolmogorovSmirnov, Chi-square, CramerVonMises, FisherExact, Least-Squares Density Difference, Z-test, Anderson-Darling test, G-test, T-test, Emperical-MMD, Total-Variation Distance, Mann-Whitney U-rank test. Team Size: 1
Company: Nexcel Solutions
My position: Junior ML Engineer
Text-Video Retrieval Github Jul 2023 – Oct 2023
My responsibilities
• Establish system pipelines and oversee the team’s work progress.
• Extract features from videos and retrieve corresponding text data to establish correlations within the database.
• Utilize a vector search index to identify matching features through cosine similarity.
• Build a web server retrieving from short video, image, text in order to gain the top-k relevant images and the suitable time interval appearing in the original videos.
• Additionally, text from video frames will be extracted using Automatic Speech Recognition (ASR), Image Captioning, and Optical Character Recognition (OCR) to generate features and compare them with the input text.” Technology Description
• Vector Database: Faiss, Elasticsearch.
• Frameworks: CLIP, PLIP, BLIP, FLIP, BERT, Transformer, TransNet, AutoShot, CRAFT, VietOCR.
• Frontend: FastAPI, JavaScript, HTML, CSS.
Team Size: 5
Competition: HCMC AI Challenge 2023
My position: Leader
Smart Menu Application Github Feb 2023 – Jun 2023 My responsibilities
• Develop a smart mobile phone using Java, train 4 main modules: Menu Scanner (using PaddleOCR built C++ core to compress native library and load to mobile phone by Java tested on DB for text detection, CRNN for text recognition, and MobileNetv3 for text direction correction).
• Machine Translation (RetNet, Seq2Seq, Transformer and LSTM) from HuggingFace.
• Develop a recommendation system using NeuMF and Collaborative Filtering
• Chatbot assistance using the RASA framework, tested with BERT and RoBERTa models. Package the images with Docker and deploy the API using the FastAPI framework on Google Cloud Platform. Technology Description
• Mobile: Java, Flutter.
• Model: DB, CRNN, MobileNetv3, RetNet, Seq2Seq, Transformer, LSTM, NeuMF, Collaborative Filtering, BERT, RoBERTa.
• Cloud: Google Cloud Platform, AWS, Heroku.
• Database: MySQL.
Team Size: 2
Subject: Graduation Project
My position: Leader
ID Card Verification - Unicloud eKYC service Github Apr 2023 – Dec 2023 My responsibilities
• Improving ID Card Verification in eKYC service by adding Correcting Image Orientation without using models to enhance time inference and accuracy.
• Detecting 4 corners on ID Card to get polygon used YOLOv5, v7, v8.
• Cutting region of interest based on perspective transform and calculating the angle to correct image orientation.
• Cutting region of interest based on perspective transform and calculating the angle to correct image orientation.
• The API of the system is built by FastAPI, packaged by Docker, and submitted to GCP to request.
• Processing ID Card data for AI Leader.
• Labeling and testing information extraction on ID Card using PaddleOCR. Technology Description
• Tools: Jira, Gitlab, Label Studio.
• Model: YOLOv5, v7, v8, DB, CRNN.
• Database: MongoDB.
Team Size: 5
Customer: Unicloud eKYC, Unicloud CA
My position: AI Engineer
Autonomous Car using Deep Learning and Computer Vision Github Sep 2022 – Nov 2022 My responsibilities
• UNET 3+, BiseNet Lane segmentation-based PID controller.
• YOLOv4-tiny, v5, v7, v8 for traffic signs detection, CNN for traffic signs recognition, and image processing.
• Simulating on Unity to test algorithms and embedding on Jetson Nano. Technology Description
• Model: YOLOv4-tiny, v5, v7, v8, UNET 3+, BiseNet.
• Techniques: PID controller, Canny, Gaussian Blur, Hough Line Transform.
• Device: Jetson Nano.
Team Size: 4
Competition: UIT Racing Car 2023.
My position: Leader
Stock Price Prediction Github May 2022 – Jul 2023 My responsibilities
• Suggesting actions and computing profit using Deep Reinforcement Learning.
• Packaging by Docker and storing API by Flask framework on Google Cloud Platform. Technology Description
• Model: LSTM, BiLSTM, CNN2D+LSTM, Deep Reinforcement Learning.
• Cloud: Google Cloud Platform.
• Frontend: Flask, JavaScript.
• Database: MongoDB.
Team Size: 4
Subject: Capstone Design Project
My position: Leader
Speech Recognition Github May 2022 – Jul 2023
My responsibilities
• Using MFCC for processing analogue signals and sampling from discrete signals by Fast Fourier Transform.
• CNN for classifying speeches, converting to tinyML(tflite), defining GPIO and embedding Raspberry Pi 4. Technology Description
• Device: Raspberry Pi 4.
• Model: CNN, Fast Fourier Transform, MFCC.
Team Size: 3
Customer: Icon technic
My position: AI R&D
Certifications
IELTS 7.0: Reading 7.0, Listening 7.5, Writing 6.5, Speaking 6.0 Link AWS Solution Architect - Associate Link
Deep Learning with PyTorch : Siamese Network Link Convolutional Neural Networks Link
Deep Learning with PyTorch : Generative Adversarial Network Link Deep Learning with PyTorch : Neural Style Transfer Link Deep Learning with PyTorch : Object Localization Link Deep Learning with PyTorch : GradCAM Link
Facial Expression Recognition with PyTorch Link
Aerial Image Segmentation with PyTorch Link
Deep Learning with PyTorch : Image Segmentation Link Honors And Awards
Graduated Valedictorian HCMUTE K19, majoring in Computer Engineering. Link Top 1 UIT Car Racing using Digital Image Processing Top 10/40 HCMC AIC 2023
Contact this candidate