Home About Projects Skills Certs Blog Contact
Available for Opportunities

ALEX
CHEN

AI Engineer  |  ML Researcher  |  Data Storyteller

Building intelligent systems that learn, adapt, and solve complex real-world problems. Specialising in deep learning, NLP, and computer vision at the intersection of research and production.

0+
Years Experience
0+
Models Deployed
0+
Research Papers
0K+
GitHub Stars
SCROLL
🧠
6+Years in AI/ML

About Me

Turning Data Into
Intelligence

"I believe AI's greatest power lies not in replacing human thought, but in amplifying it. Every model I build is a step toward a world where technology serves humanity more meaningfully."
Location
San Francisco, CA
Role
Senior AI Engineer
Focus
LLMs & Computer Vision
Status
● Open to Work

Core Expertise

Deep Learning NLP / LLMs Computer Vision Reinforcement Learning Generative AI TensorFlow PyTorch Hugging Face OpenAI API AWS SageMaker GCP Vertex AI Docker / K8s

Portfolio

Featured Projects

Production-grade AI systems, research prototypes, and open-source contributions spanning NLP, computer vision, and beyond.

🤖
LLMRAGFastAPI
Enterprise RAG Pipeline

Built a production RAG system for a Fortune 500 company, enabling semantic search over 2M+ internal documents with sub-second retrieval.

↑ 94% accuracy · ↓ 60% hallucination rate · 2M+ docs indexed
👁️
CVYOLOv8ONNX
Real-Time Defect Detection

Computer vision system for a semiconductor fab detecting nanoscale defects at 120 FPS on edge hardware using a custom-pruned YOLOv8 architecture.

↑ 99.2% precision · 120 FPS on edge · $2M annual savings
💬
NLPTransformersRLHF
Domain-Adapted LLM Fine-tuning

Fine-tuned LLaMA-3 on legal corpora using QLoRA + RLHF, achieving GPT-4 parity on legal reasoning benchmarks at 1/20th the inference cost.

↑ GPT-4 parity · ↓ 95% inference cost · ICLR 2024 workshop
📈
RLTime SeriesPyTorch
Algorithmic Trading Agent

Deep RL agent using PPO + Transformer-based market encoder for intraday equity trading, consistently outperforming S&P 500 in backtests.

↑ 34% annualised return · Sharpe 2.1 · Live paper trading
🗣️
ASRTTSStreaming
Multilingual Voice Assistant

End-to-end voice AI supporting 22 languages with <200ms latency, built on Whisper + custom TTS with prosody control via diffusion model.

22 languages · <200ms · 4.8★ app store rating
🧬
BioAIGNNDrug Discovery
Protein Binding Predictor

Graph Neural Network predicting protein-ligand binding affinity for drug discovery, reducing wet-lab screening cost by 80% for a pharma partner.

RMSE 1.2 kcal/mol · 4× faster screening · NeurIPS 2023

Expertise

Skills & Tools

A breadth of technical capabilities from research to production deployment.

Languages
Python98%
R / Julia82%
SQL / NoSQL88%
C++ / CUDA70%
ML Frameworks
PyTorch96%
TensorFlow / Keras91%
Hugging Face95%
scikit-learn93%
Cloud & MLOps
AWS SageMaker88%
GCP Vertex AI82%
Docker / Kubernetes85%
MLflow / W&B90%
AI Specialisation Radar
🔥PyTorch
🤗HuggingFace
🧠LangChain
☁️AWS
🐳Docker
📊W&B
CUDA
🔗LlamaIndex
🚀FastAPI

Background

Education & Certifications

Academic foundation and industry credentials that underpin my practice.

🎓
Ph.D. Computer Science
Stanford University
2018 – 2022 · GPA 3.98/4.0
📐
B.Sc. Mathematics & CS
MIT
2014 – 2018 · Summa Cum Laude
🏆
Kaggle Grandmaster
Kaggle · Top 0.1%
2020 · Rank #42 Global
Google Professional ML Engineer2024
Google Cloud
Advanced certification covering ML model design, training, operationalisation, and MLOps on Google Cloud Platform.
AWS Certified ML Specialty2023
Amazon Web Services
Speciality cert validating expertise in designing, implementing, and maintaining ML solutions on AWS infrastructure including SageMaker.
Deep Learning Specialization2022
DeepLearning.AI · Andrew Ng
5-course series covering CNNs, RNNs, Transformers, and practical deep learning engineering from Professor Ng himself.
NVIDIA DLI – Generative AI2023
NVIDIA Deep Learning Institute
Hands-on training in diffusion models, LLM fine-tuning, and deploying generative AI at scale with Triton Inference Server.

Insights

From the Lab

Deep dives into AI research, engineering patterns, and experiments from the frontier of machine intelligence.

LLMsMar 2025
Why RAG Still Beats Fine-Tuning for Most Enterprise Use Cases

After deploying both approaches at scale, here's what the benchmarks don't tell you about retrieval-augmented generation vs. domain fine-tuning.

Read article →
MLOpsFeb 2025
The Hidden Cost of ML Technical Debt: A Production Autopsy

How one data leakage bug survived 6 months of code review, deployed to 12 markets, and quietly degraded model performance by 18%.

Read article →
ResearchJan 2025
QLoRA in Practice: Fine-Tuning a 70B Model on a Single A100

A practical walkthrough of quantised LoRA fine-tuning with actual throughput numbers, memory traces, and lessons learned from three failed attempts.

Read article →

Get In Touch

Let's Build
Something

Whether it's a research collaboration, consulting inquiry, or an ambitious AI product — I'd love to hear about it.

⬇ Download Resume