AI

NGUYEN HUYNH HOANG KHA

AI Engineer

Open
Hire Me

AI ENGINEER BRAND LAUNCH

Applied AI Systems Interface

Loading...

Syncing knowledge layer

0%

NLPComputer VisionRAGMLOpsMultimodal

AI Engineer • Tan Binh District, Ho Chi Minh City

A Creative

AI ENGINEER

Engineering Applied AI Systems from Research Signals to Production Impact

AI Engineer specializing in NLP, Computer Vision, Generative AI, and production-grade deployment for enterprise chatbots, reasoning assistants, and multimodal products.

From model experimentation to deployment, I build AI systems that solve real business and learning problems with measurable technical outcomes across retrieval quality, latency, and product usability.

View ProjectsDownload CVContact Me

Available for applied AI engineering and product-focused R&D roles

Scope

3+ AI Domains

Stack

RAG + Multimodal

Recognition

Top 10 GCSB

Outcome

+40% Retrieval Gain

PortfolioEmail
Multimodal AIRAG SystemsNLP + CV + GenAIModel DeploymentProduction-Minded AI Engineering
NLPComputer VisionGenerative AIRAG SystemsMulti-hop ReasoningMLOps / Deployment

About

An applied AI engineer connecting research ideas to deployable systems

I work across model development, retrieval architecture, and deployment pipelines to transform AI concepts into usable products.

AI Engineer with hands-on implementation experience in NLP, Computer Vision, and Generative AI, focused on turning advanced models into practical systems.

My work spans LLM pipelines, multimodal retrieval, reasoning-oriented QA, and AI chatbot products designed for enterprise and small-business workflows.

I combine research curiosity with engineering discipline: evaluate rigorously, tune for real constraints, and integrate reliably into production-facing environments.

Applied AI Scope

NLP, Computer Vision, Generative AI, multimodal retrieval, and reasoning-oriented assistants.

Engineering Method

Build, evaluate, tune, and integrate into production-facing workflows with measurable priorities.

Practical Reliability

Focus on deployment constraints, latency, retrieval quality, and maintainable architecture decisions.

Cross-Context Delivery

Experience spanning enterprise systems, educational assistants, and business automation use cases.

Experience

Production-focused roles across enterprise delivery, freelancing, and AI instruction

A timeline centered on system implementation, deployment quality, and practical AI adoption in business and education settings.

CYBERSOFT TECHNOLOGY CO.,LTD

AI Engineer

Feb 2024 – Nov 2025
  • Built and deployed AI models for enterprise chatbot and educational assistant use cases in real-world environments.
  • Improved production readiness through model evaluation, fine-tuning, and deployment integration across business scenarios.
  • Developed agentic AI teaching materials and demo projects for internal adoption and corporate training programs.
  • Collaborated across technical and non-technical stakeholders to translate AI capabilities into usable product workflows.

Freelance

AI Engineer

Nov 2022 – Jan 2023
  • Delivered custom AI chatbot solutions for small businesses to automate customer support and service operations.
  • Integrated AI systems into workflow automation for data processing, internal communication, and task management.
  • Researched and implemented NLP + LLM applications for financial trading analysis and news interpretation.
  • Designed practical solution scopes that balanced speed-to-value, model performance, and maintainability.

Education & Training

Mentor DA / AI Instructor

Jun 2025 – Present
  • Taught Power BI and SQL for data analytics foundations and business reporting workflows.
  • Instructed Machine Learning and Deep Learning concepts with practical implementation orientation.
  • Supported learner progression from core concepts to applied problem-solving in analytics and AI.

Featured Projects

Systems that prove depth across multimodal retrieval, reasoning QA, and computer vision

A curated project stack showing architecture thinking, implementation rigor, and measurable outcomes.

Multimodal Video Q&A System project thumbnail
Role · Lead AI Engineer

RAG-based chatbot that understands video, text, image, and speech in one retrieval workflow.

Multimodal Video Q&A System

Engineered an interactive multimodal retrieval assistant for e-learning and knowledge management, combining modality-specific understanding with RAG orchestration.

Problem

Knowledge in video, speech, image, and text was fragmented across separate retrieval paths.

Solution

Built a unified multimodal RAG workflow with Whisper, CLIP, transformer embeddings, and LangGraph orchestration.

Impact

Improved retrieval accuracy by 40% and reduced latency by 30% for an interactive e-learning prototype.

Technical Highlights

  • Integrated Whisper, CLIP, and transformer embeddings for multimodal indexing and retrieval.
  • Orchestrated retrieval and reasoning flow using LangChain, LangGraph, FAISS, and Chroma.
  • Implemented interactive interface with Gradio and OpenAI API tooling.

Outcomes

  • Improved retrieval accuracy by 40%.
  • Reduced end-to-end latency by 30%.
  • Delivered a working prototype for educational and knowledge-centric use cases.
WhisperCLIPTransformersFAISSChromaLangChainLangGraphOpenAI APIsGradio
Multimodal AIRAGE-LearningApplied GenAI
Links coming soon
Multi-Hop Reasoning QA Chatbot project thumbnail
Role · AI Engineer

MVP QA engine using graph-aware multi-hop reasoning over DBLP and Wikidata.

Multi-Hop Reasoning QA Chatbot

Built a question-answering MVP that decomposes complex queries, performs graph-informed relation traversal, and surfaces multi-step answers via lightweight interfaces.

Problem

Complex questions required multi-step reasoning across DBLP and Wikidata relations.

Solution

Designed decomposition + graph-aware QA flow with LLaMA-3/Qwen-7B, GraphSAGE/GAT, and cache-backed APIs.

Impact

Validated a production-lean MVP with faster repeated-query handling via SQLite/FAISS caching.

Technical Highlights

  • Used LLaMA-3 and Qwen-7B for question decomposition and reasoning path planning.
  • Applied GraphSAGE and GAT for multi-step relation encoding across knowledge sources.
  • Served system via FastAPI + Streamlit with caching using SQLite and FAISS.

Outcomes

  • Validated feasibility of multi-hop QA for structured and semi-structured knowledge retrieval.
  • Reduced repeated query overhead with cache-backed retrieval primitives.
LLaMA-3Qwen-7BGraphSAGEGATFastAPIStreamlitSQLiteFAISS
Reasoning QAKnowledge GraphMVPLLM Systems
Links coming soon
MHQ-REACTRAG project thumbnail
Role · AI Engineer & Research Builder

Research-grade assistant combining Chain-of-Thought, ReAct loops, and RAG with verification.

MHQ-REACTRAG

Designed a multi-hop QA assistant architecture that blends tool-augmented reasoning, retrieval pipelines, and verification layers to reduce hallucinations in complex tasks.

Problem

Multi-hop assistants can hallucinate when retrieval and reasoning are weakly coupled.

Solution

Combined Chain-of-Thought, ReAct loops, tool-use, and FEVER-style verification in a research-grade architecture.

Impact

Raised reliability and evaluation rigor with reproducible benchmarks, logging, and ablation-ready experiments.

Technical Highlights

  • Integrated tool use across web search, vector retrievers, Neo4j knowledge graph, and Python sandbox.
  • Applied FEVER-style verification loop to improve answer reliability.
  • Built reproducible experiment workflows with logging, benchmarks, and ablation studies.
  • Supported deployment with FastAPI backend and Streamlit/Gradio frontends.

Outcomes

  • Delivered an extensible research-to-application architecture for robust multi-hop QA.
  • Enabled benchmark-driven iteration for reliability and system quality.
CoTReActRAGFastAPIStreamlitGradioNeo4jFAISSMilvusPinecone
Research EngineeringMulti-hop QATool UseVerification
Links coming soon
Virtual Try-On 2D project thumbnail
Role · Computer Vision Engineer

Computer vision pipeline for realistic digital garment fitting in e-commerce scenarios.

Virtual Try-On 2D

Implemented a 2D virtual try-on system that maps garments onto user photos using a staged visual pipeline for practical e-commerce prototyping.

Problem

E-commerce try-on flows often fail to show realistic garment fit from user photos.

Solution

Implemented a 2D try-on pipeline with human parsing, pose estimation, U-Net, and geometric image warping.

Impact

Delivered realistic garment visualization from user-photo + clothing-image inputs for shopping use cases.

Technical Highlights

  • Combined human parsing, pose estimation, U-Net segmentation, and image warping.
  • Optimized the transformation flow to preserve garment structure and body alignment.
  • Produced realistic visual previews from user-photo and garment-image inputs.

Outcomes

  • Built a working visualization pipeline for applied e-commerce use cases.
  • Demonstrated practical CV integration for user-facing digital experiences.
OpenCVPose EstimationHuman ParsingU-NetImage Warping
Computer VisionE-commerceImage SynthesisApplied AI
Links coming soon
Student’s Behaviour Detection project thumbnail
Role · Computer Vision Engineer

Real-time classroom engagement detection with behavioral signals and time-series analysis.

Student’s Behaviour Detection

Developed a monitoring system that captures in-class behavioral patterns and infers attention and performance-related outcomes from live visual signals.

Technical Highlights

  • Used YOLOv8 and MediaPipe for real-time signal extraction.
  • Modeled behavioral events through temporal patterns and engagement markers.
  • Connected detection outputs to performance and attention-related predictions.

Outcomes

  • Delivered a classroom-focused engagement analytics prototype.
  • Showed feasibility of behavior-aware AI support in education settings.
YOLOv8MediaPipeTime-Series AnalysisComputer Vision
Real-time CVEdTechBehavior AnalyticsPrediction
Links coming soon

Skills

Interactive AI Techstack

An interactive skill map that surfaces core AI tooling across model development, retrieval systems, and production deployment.

MY TECHSTACK

TensorFlow
PyTorch
Scikit
XGBoost
LightGBM
Transformers
LLaMA
GPT
BART
SpaCy
NLTK
OpenCV
YOLOv8
Detectron2
MMDet
Optuna
FAISS
Pinecone
ChromaDB
Docker
K8s
MLflow
FastAPI

AI & Machine Learning Skills

  • Machine Learning & Deep Learning: TensorFlow, PyTorch, Scikit-learn, XGBoost, LightGBM
  • Natural Language Processing (NLP): Hugging Face Transformers, SpaCy, NLTK, BART, GPT, LLaMA
  • Computer Vision (CV): OpenCV, YOLOv8, Detectron2, MMDetection
  • Model Optimization & Training: Hyperparameter Tuning (Optuna, GridSearchCV), Bayesian Optimization

AI Deployment & MLOps

  • MLOps & Model Deployment: Docker, Kubernetes, CI/CD, MLflow, FastAPI, Flask
  • Vector Search & Embeddings: FAISS, Pinecone, ChromaDB, RAG Pipeline
  • AI Serving & APIs: TensorFlow Serving, Triton Inference Server, RESTful APIs, GraphQL
  • Data Processing & Feature Engineering: Pandas, NumPy, Dask, Polars

Selected Impact

Signal-rich outcomes that show both technical depth and delivery maturity

Metrics and implementation highlights grounded in project execution across multimodal AI, enterprise chatbot systems, and practical automation.

+40%

Retrieval Quality

Retrieval accuracy improvement in multimodal video Q&A workflow.

-30%

System Latency

Latency reduction in RAG chatbot inference and response pipeline.

Multi-domain

Applied Delivery

Built enterprise chatbots, educational assistants, and business workflow automations.

Top 10

Recognition

Recognized as a top 10 AI Developer in Google Cloud Skill Boost.

Multimodal System Engineering

Delivered unified AI experiences across text, image, speech, and video retrieval interfaces.

Reasoning-Oriented QA

Built multi-hop, tool-augmented QA assistants with verification flows to improve reliability.

Deployment & Integration

Integrated models into FastAPI services, interactive frontends, and business-facing workflows.

Training & Enablement

Created teaching programs and demos that accelerate AI adoption within teams and learning contexts.

Publication

Research-backed perspective on modern RAG architecture design

This publication highlights system-level thinking across orchestration frameworks, vector stores, and open-source LLM integration choices.

Retrieval-Augmented Generation Architectures: Integrating LangChain, LangGraph, Vector Stores, and Open-Source Large Language Models — A Survey and Case Study with GPT-OSS-20B

Author: Kha Nguyen

A survey-and-case-study publication that examines modern RAG architecture decisions across orchestration frameworks, vector stores, and open-source LLM deployment pathways.

Publication Link Placeholder

Recognition

Focused validation from Google AI learning and challenge ecosystems

Selected recognitions and credentials that reinforce practical execution strength in applied AI engineering.

Google Cloud Skill Boost — Top 10 AI Developer

Recognized for strong performance in AI-focused international challenge tracks.

Google Machine Learning & Generative AI Certificates

Holds multiple Google certificates and badges across ML and GenAI learning pathways.

Certificates

14 Basics Certificates and Badges and 1 Intermediate Certificate about Machine Learning and Generative AI from GOOGLE

Complete certificate archive from Google Cloud Skills Boost, covering responsible AI, LLM foundations, MLOps, multimodal RAG, and vision-generation topics.

View Profile
  • Introduction to Responsible AI
  • Introduction to Large Language Models
  • Introduction to Generative AI
  • Responsible AI for Developers: Interpretability & Transparency
  • Responsible AI for Developers: Fairness & Bias
  • Machine Learning Operations (MLOps) for Generative AI
  • Inspect Rich Documents with Gemini Multimodality and Multimodal RAG
  • Vector Search and Embeddings
  • Introduction to Vertex AI Studio
  • Create Image Captioning Models
  • Transformer Models and BERT Model
  • Encoder-Decoder Architecture
  • Attention Mechanism
  • Introduction to Image Generation

Education

Formal computer science foundation

Grounded in core computer science training and extended through practical AI engineering work.

Bachelor in Computer Science

University of Information and Technology

Graduation year is not specified in the provided CV details.

Contact

Let’s build intelligent systems that create real business and product impact

Open to AI Engineer, Applied AI, GenAI, and intelligent product opportunities across startup and product-focused engineering teams.

Open to Work
Open to AI Engineer, Applied AI, GenAI, and intelligent product opportunities.
Open to Work< 24h ResponseAI Engineer / Applied AI / GenAI

Usually responds within 24 hours for hiring and collaboration requests.

Available for full-time roles, product-focused R&D, and high-impact AI delivery.

Email

nhhkha.91tn@gmail.com

Phone

0769532711

Website

www.khaportfolio.info

Social

Open to Work: Contact NowReview Full CV