Available for AI / ML engineering roles

Jia Xiang Lim

Senior Applied AI Scientist · Machine Learning Engineer

Building production-grade LLM, recommendation, and retrieval systems — fine-tuning, RAG, and agentic AI — with 5+ years taking models from research to scale.

5+Years in production ML/AI
1B+Annual inference requests
1Granted US patent
About

Turning AI from notebook to production

How I approach machine learning — rigorous, measurable, and built to scale.

I'm a Senior Applied AI Scientist & Machine Learning Engineer specializing in LLM systems, recommendation, and retrieval, with 5+ years building and deploying production-grade ML. I have a proven track record of improving search relevance, personalization, and operational efficiency through large-scale recommendation models and retrieval pipelines.

I design end-to-end LLM pipelines — fine-tuning (Mistral-7B, E5-Instruct), RAG systems, and multi-agent architectures — alongside scalable evaluation and experimentation frameworks that drive measurable business impact. I'm a co-inventor of a granted US patent (No. 12,450,268) and second author of a SIGIR 2026 paper.

LLM Systems

End-to-end LLM pipelines — fine-tuning (Mistral-7B, E5-Instruct), RAG, and multi-agent architectures.

Recommendation & Retrieval

Large-scale recommendation and semantic retrieval/reranking pipelines serving 1B+ inference requests annually.

Model Fine-Tuning

PEFT / LoRA fine-tuning that cut compute costs ~40% while maintaining model performance.

Agentic AI

Multi-agent systems (LangGraph, CrewAI) for planning, retrieval, reasoning, and execution.

MLOps

Centralized experimentation and model lifecycle infra with MLflow on Kubernetes (Kyma).

Research & Patents

Granted US Patent No. 12,450,268, SIGIR 2026 co-author, and presenter at NVIDIA GTC 2025.

Skills

A full-stack AI toolkit

From model training to data pipelines to cloud infrastructure — the stack behind production AI.

GenAI / LLM

Production LLM systems — fine-tuning, retrieval-augmented generation, evaluation, and agents that act.

LangChainLangGraphLlamaIndexCrewAIRAGMulti-Agent SystemsPEFT / LoRAAnthropic Claude

Machine Learning & AI

Deep learning, recommendation, and information retrieval.

PyTorchTensorFlowHuggingFaceNVIDIA NeMoRecommender SystemsInformation Retrieval

Data & Databases

Modeling & querying at scale.

PostgreSQLMySQLNeo4j

MLOps / Cloud

Reproducible, cloud-native ML.

AWSDockerKubernetesTerraformMLflowJenkins

Languages & Backend

From data pipelines to production services and APIs.

PythonJavaJavaScriptFastAPISpring BootFlask
Experience

A track record of shipping

Roles where I owned ML and AI systems end-to-end — from framing to production.

Jul 2022 — Present

SAP

Senior Applied AI Scientist / Machine Learning Engineer

SAP

Jul 2022 — PresentSingapore · Recommendation Team

Building large-scale recommendation, retrieval, and LLM systems for enterprise search and personalization.

  • Optimized large-scale transformer-based sequential recommendation systems supporting over 1B online and offline inference requests annually — improving inference latency, recommendation quality, and system throughput.
  • Fine-tuned and optimized LLMs (Mistral-7B, E5-Instruct-Large) for recommendation and semantic retrieval, reducing compute costs by ~40% through PEFT/LoRA while maintaining model performance.
  • Designed and productionized a transformer-based query suggestion system to improve query understanding and response latency, resulting in granted US Patent No. 12,450,268.
  • Built production LLM evaluation frameworks for RAG (LLM-as-a-Judge, pairwise ranking, rubric-based scoring), reducing evaluation time by ~50% — presented at NVIDIA GTC 2025.
  • Architected centralized ML experimentation and model lifecycle infrastructure using MLflow on Kubernetes (Kyma).
  • Second-authored “RecPFN: Prior-Fitted Networks for In-Context-Based Recommendations,” accepted at SIGIR 2026.
PyTorchMistral-7BPEFT / LoRARAGMLflowKubernetes

Jul 2021 — Jul 2022

DBS Bank

AI Developer

DBS Bank

Jul 2021 — Jul 2022Singapore · Conversational AI

Improved conversational AI quality and built knowledge-graph-powered NLP services.

  • Improved chatbot intent detection and response accuracy by 2–3% through NLP model optimization and iterative experimentation.
  • Built and deployed Named Entity Recognition (NER) pipelines using SpaCy and rule-based approaches for structured information extraction.
  • Designed and managed a scalable knowledge graph (Neo4j) to model entity relationships and enable efficient querying.
  • Developed graph-based recommendation systems (PageRank, Jaccard similarity), improving recommendation accuracy by 2–3%.
  • Engineered backend services using Java Spring Boot and Flask, supporting scalable ML-powered applications.
PythonSpaCyNeo4jSpring BootFlask

Aug 2019 — Jul 2021

ST Engineering

System Safety Engineer

ST Engineering

Aug 2019 — Jul 2021Singapore

Led system-level risk analysis and safety assurance across the full product lifecycle.

  • Conducted system-level risk analysis using Hazard Analysis and Fault Tree Analysis (FTA) to identify failure points and mitigate risks.
  • Developed and presented comprehensive safety cases to internal and external stakeholders, supporting design validation and regulatory approval.
  • Led system safety initiatives across the full product lifecycle, partnering with project management to enforce safety standards and compliance.
Hazard AnalysisFault Tree AnalysisSafety Cases
Featured Projects

Systems I've designed & shipped

Selected work spanning agentic platforms, automated trading, and applied GenAI.

Featured project

OpenClaw

Automated algorithmic trading platform

An automated algorithmic trading platform integrated with Interactive Brokers — combining agent workflows, risk controls, and execution automation in one system.

  • Portfolio monitoring
  • Risk management
  • Agent workflows
  • Execution automation
  • Trade analytics
PythonLangGraphFastAPIPostgreSQLInteractive Brokers API
In active development
Architecture diagram
Demo video
In active development

ArchMind

AI software architect

An agentic platform that reasons through system design end-to-end — capturing requirements, evaluating trade-offs, and producing architecture diagrams.

  • Requirements capture
  • Trade-off analysis
  • Architecture diagrams
  • Multi-agent reasoning
LangGraphCrewAIRAGFastAPIAWS
In active development

AI Travel Planner

Personalized itinerary generation

An AI-powered travel planning system that generates personalized itineraries, budgeting recommendations, and tailored travel suggestions.

  • Personalized itineraries
  • Budget recommendations
  • Travel suggestions
  • Conversational UI
LangChainRAGNext.jsPostgreSQL
Contact

Let's build something intelligent

Open to Applied AI Scientist / ML Engineer roles and ambitious LLM, recommendation, and agentic AI projects. Reach out through any channel below.