Ivan Koh

I'm Ivan, ML/AI engineer for investment teams.

I'm a forward-deployed ML/AI engineer and data scientist focused on private markets. I turn investor theses into production data products and ML services. Leading technical initiatives across portfolio companies - from demand forecasting for startups to real-time investment intelligence platforms. My work spans hands-on data science, production engineering, and everything in between.

Explore case studiesContact me

Featured Work

Private Markets Investment Platform - Backend, Data, and AI Engineering

Challenge

Translate investor theses into scalable, production data products while maintaining reliability and speed across internal intelligence platforms.

Approach

Led all data/ML initiatives; owned firm-wide GCP with an emphasis on performance, cost, and scale; standardized CI/CD and observability.

Solution

Shipped production services and real-time pipelines powering portfolio monitoring, deal workflows, and investment intelligence.

Impact

Supported 55+ portfolio companies; High availability; -70% analyst time.

Python
GCP
BigQuery
Terraform
DevOps
MLflow

Forward-Deployed ML/AI for Portfolio Success

Challenge

Portfolio companies required targeted ML/AI solutions to drive measurable outcomes: demand planning, personalization, and workflow automation.

Approach

Embedded as FDE/Data Scientist with each company to build tailored ML solutions within their context and constraints.

Solution

Demand forecasting for SEA's largest D2C womenswear retailer ($88M+ revenue); personalized recommendations for Indonesia's biggest coffee chain (1,100+ outlets); automated prompt optimization (DSPy) for a healthtech platform (20M+ users).

Impact

Built backbone for robust forecasting, improve customer engagement, and accelerated healthcare workflows with a focus on performance and cost.

Python
PyTorch
XGBoost
DSPy
OpenAI
FastAPI

Infrastructure That Powers Investments

Challenge

Investment operations required reliable, secure, and scalable data infrastructure to support real-time decision-making across multiple teams and portfolio companies.

Approach

Led DevOps efforts focusing on system reliability and secure data operations, while building infrastructure that could scale with the firm's growing portfolio.

Solution

Designed and implemented production-grade infrastructure on GCP with automated CI/CD, real-time processing, and secure multi-tenant architecture across internal platforms.

Impact

99.9% uptime, secure data handling, and faster delivery cycles for investment tools.

GCP
Docker
CI/CD
PostgreSQL
Terraform
DevOps

Projects

Startup Funding Success Prediction

Developed statistical models and ML algorithms to predict startup funding success, integrating multiple data sources including web traffic and founder profiles.

Python
PyTorch
XGBoost
MLflow
Machine Learning

LLM-Powered Visual Document Understanding

Built an end-to-end system combining fine-tuned YOLO models with LLMs for financial document analysis and investment due diligence automation.

PyTorch
HuggingFace
RAG
LLM
Deep Learning

Image Captioning with Deep Learning

Built encoder-decoder models with attention to generate descriptive image captions. Used PyTorch and Weights & Biases for experiment tracking.

PyTorch
HuggingFace
W&B
Deep Learning

LLM-Powered Data Analyst

Developed a RAG system for Singapore-specific data queries. Created a vector database with FAISS for efficient similarity search.

Langchain
FAISS
Streamlit
RAG
LLM

Housing MLOps Pipeline

Built end-to-end ML pipeline with Dockerized Airflow workflows for data processing and MLFlow for model versioning and tracking.

Docker
Airflow
MLFlow
MLOps

Kickstarter Campaign Success Predictor

Built a model predicting Kickstarter campaign success to help backers optimize investments by analyzing historical data to identify key success factors.

Python
Scikit-learn
Data Analysis
Machine Learning

Let's Connect

Open to discussing ML systems, investment intelligence, or the intersection of technology and finance.