Profile
I'm Tien Cheng, an undergraduate reading Computer Science at the National University of Singapore (NUS) on the NUS Global Merit Scholarship. I'm a graduate of Singapore Polytechnic (SP) with a Diploma in Applied AI and Analytics, and a recipient of the Singapore Computer Society Silver Medal. As a former co-founder and operations department head of SPAI (an AI Singapore Student Chapter) in AY21/22, I actively promoted the field of Data Science during my time at SP.
I completed my full-time national service as a Software Developer at the Digital and Intelligence Service, and I'm currently a co-founder of Conform Labs. Outside of work and school, I serve as President of RC4Entre and a Coreteam member of NUS Hackers, and I also work on personal projects and competitions in my free time.
Education
B.Comp. in Computer Science
Expected 2029 at National University of Singapore, Singapore- NUS Global Merit Scholarship.
- GPA: 4.79/5.0.
Diploma in Applied AI and Analytics
2020 to 2023 at Singapore Polytechnic, Singapore- GPA: 4/4
- Relevant Coursework: Front End Web Development, Back End Web Development, AI and Machine Learning, Deep Learning, DevOps for Automa‑ tion and AI, Programming for Data Analytics and Statistics for Data Science
- Co-founded SPAI, an AI Singapore Student Chapter, organising three successful data science workshops (attended by over 80 students), helping SPAI to win an SP Excellence Award in its first year.
Experience
Co-Founder at Conform Labs
From August 2025 to December 2025 at Conform Labs, Singapore- Built a voice-agent evaluation framework to compare agent quality across simulated conversations using automated scoring and Langfuse trace analysis.
- Developed form analytics, document-to-form generation, and real-time voice-agent pipelines with TypeScript, FastAPI, and Pipecat.
Software Developer at the Digital and Intelligence Service, Singapore Armed Forces (National Service)
From May 2023 to August 2025 at Singapore Armed Forces, Singapore- Attained the platoon best trainee award in the Signals Institute vocational training course.
- Built an LLM-powered exam-marking platform that reduced instructor grading time by over 50% in an air-gapped environment.
- Improved grader accuracy and robustness by benchmarking prompts and automating prompt optimization with DSPy and open-source data.
- Integrated Langfuse telemetry to monitor traces, latency, and scoring quality, reducing time spent debugging the pipeline.
Co-Founder at Flora Softworks
From February 2023 to October 2023 at Flora Softworks, Singapore- Flora is a tech startup that builds software to help organisations manage their information, with our primary product being an enterprise search application.
- Built a neural search API for unstructured enterprise documents using Elasticsearch, ONNX Runtime, Jina, FastAPI, and MongoDB, supporting public-sector and enterprise pilots.
- Developed unit tests for the AI microservices using Pytest, and also set up a Kubeflow pipeline on Vertex AI for running evaluation of our neural search engine using BEIR.
- Improved document-layout detection throughput by 400% by distilling and quantizing an RTMDet model for production inference, while also improving mean average precision (mAP) from 0.41 to 0.5 (a 21.95% improvement).
- Deployed pilot systems on Google Cloud Platform and refactored services for self-hosted environments to enable on-premise adoption with clients including SkillsFuture SG, NSCC, and Coherent, Inc.
- Used OpenTelemetry in combination with Prometheus and Grafana to track and monitor the performance of the application, helping to identify and resolve memory leaks and bottlenecks.
Intern at Defence Science and Technology Agency
From September 2022 to February 2023 at Defence Science and Technology Agency, Singapore- Core developer of the AI App Store project, an in-house platform built using Vue.js and FastAPI for AI developers to publish and share their models in the form of model cards, and deploy containers as KNative services through a GUI interface.
- Developed API endpoints to enable the managing of services in a Kubernetes cluster through the app, as well as integration with the ClearML platform.
- Developed a custom WYSIWYG editor using the TipTap library, enabling the insertion of editable Plotly.js graphs within the markup.
- Wrote and maintained unit tests, and established automated testing through a CI pipeline deployed on GitHub Actions, resulting in a more reliable and robust application.
- Set up Helm charts, documentation and deployed the project on a Kubernetes cluster for use internally by two departments within DSTA.
- Contributed to a separate team project where I developed microservices for a prototype real time video analytics pipeline, integrating models like YOLOv5 and ByteTrack, using PyTorch, OpenCV, Triton Inference Server and Apache Kafka.
Leadership
President at RC4Entre
AY26/27 at RC4Entre, National University of Singapore- Leading RC4Entre, the entrepreneurship CCA of Residential College 4 at NUS.
Coreteam Member at NUS Hackers
From 2026 at NUS Hackers, Singapore- Member of the Coreteam, the organising group behind NUS Hackers' events and initiatives for the student hacker community.
Certifications
- Generative AI with Large Language Models (2023)
- Introduction to Machine Learning in Production (2022)
- Deep Learning Specialization (2022)
- Machine Learning (2021)
- Wiley Certified Data Analyst (2020)
Skills
- Python
- TypeScript
- SQL
- PyTorch
- Hugging Face Transformers
- DSPy
- RAG
- ONNX Runtime
- Scikit-learn
- FastAPI
- Streamlit
- Celery
- PostgreSQL
- PGVector
- MongoDB
- Elasticsearch
- Docker
- Kubernetes
- Google Cloud Platform
- Langfuse
- OpenTelemetry
Awards
- NUS Global Merit Scholarship (2023)
- Singapore Computer Society Silver Medal (2023)
- Best Junior Hack of the NTU MLDA Deep Learning Week Hackathon (2021)
- Champion of the Collegial Artificial Intelligence Innovation Competition (2020)
- DSTA Polytechnic Digital Scholarship (2020)
- 1st Place (University Category), HPC Innovation Competition 2023, for OwlShield, an LLM safety firewall
- Contributed a LiteLLM integration to NVIDIA's Garak, an open-source LLM security scanner (8k+ stars)