zhifeng

Zhifeng Li

Thoughtful AI. Engineered for reality.

AI/ML Engineer & Computational Scientist

I connect models, data and software to build systems people can understand and use.

Explore selected workDownload CV

Imperial College London · Applied AI & software engineering

01 / The perspective

A model is only the beginning.
The interesting work happens around it.

Which evidence should it trust? When should it ask a person? How do we know it works? From local AI research to backend services, these are the questions that shape how I build.

02 / Selected work

Ideas, made tangible.

Two ongoing explorations in making AI useful. A wider body of work in vision, learning and scientific computing.

01 — Ground the answer

Keep the intelligence local. Keep the evidence close.

Building a policy copilot that helps KYC/AML analysts find supporting evidence. Local inference, traceable citations and a human in the decision loop.

In progressKYC/AML RAG
Read case study
01Policy evidence
02Local retrieval + SLM
03Cited answer
Conceptual workflowLocal by design. Human by decision.

02 — Connect the workflow

An inbox full of noise. A next step that is clear.

Turning recruitment emails into reviewable applications, calendar events and a daily brief. The model interprets the message; deterministic rules and human review govern the next step.

In progressInbox Agent
Read case studyView source repository
01Recruitment email
02Extract + review
03A clear next step
Conceptual workflowInterpret. Review. Then act.

Different problems. The same curiosity.

Prototype / 03

ArUco + MediaPipe Hand Tracking

A computer-vision prototype combining calibrated ArUco pose estimation, MediaPipe hand landmarks and a Kalman-based fusion path with real-time visualisation and recordable experiment outputs.

  • Python
  • OpenCV
  • MediaPipe
Read case studyView source repository

Shipped / 04

EveryDay Learning AI Study App

A shipped enterprise AI/NLP workflow for the Everyday Learning app that turned saved articles into quiz-ready questions, distractors and seven-day review sessions for the Android learning experience.

  • Python
  • PyTorch
  • Hugging Face Transformers
Read case study

Coursework / 05

Predicting the Unpredictable

Team coursework exploring four storm-prediction tasks across 800 example storms, supported by a shared Python package for data, metrics, preprocessing, plotting and submission workflows.

  • Python
  • PyTorch
  • scikit-learn
Read case study

Coursework / 06

Genetic Algorithm Circuit Optimizer

A C++20 team coursework system combining circuit simulation, genetic optimisation, structural validity checks and reproducible seeded sweeps with optional robustness extensions.

  • C++20
  • CMake
  • OpenMP
Read case study

03 / How I build

From a good question to a dependable system.

01

Find the signal.

Start with the data, the user and the constraints. Make the problem precise before choosing a model.

Python · SQL · Retrieval · Data pipelines

02

Build the whole path.

Connect inference to the software around it: APIs, state, review steps and interfaces that make the result useful.

Java · FastAPI · LangGraph · Backend systems

03

Make it testable.

Use explicit baselines, reproducible experiments and regression checks. Understand where a system works and where it needs a person.

Evaluation · Pytest · C++ · Numerical methods

04 / The journey

Built across disciplines.

A computer science foundation. Experience with real software workflows. A growing focus on applied AI and computational science.

  1. Jun — Sep 2026

    Microsoft Core AI

    Research internship

    Local SLM and evidence-grounded RAG for KYC/AML policy support, connected to my Imperial MSc research. The research case study remains in progress.

  2. Jun — Sep 2025

    Kuaishou

    Sales Copilot · AI engineering

    Worked on retrieval, tool routing and enterprise question-answering workflows for a sales assistant.

  3. Nov 2023 — Apr 2024

    MHR × University of Nottingham

    EveryDay Learning · AI/NLP development

    Built article-to-quiz workflows, distractor generation and spaced review scheduling for an Android learning experience.

  4. Jun — Sep 2023

    Meituan

    Backend engineering internship

    Worked on rewards workflows and backend services, connecting business rules with reliable data processing.

Imperial College London

MSc Applied Computational Science and Engineering

2025 — 2026 · Expected award December 2026

University of Nottingham Ningbo China

BSc Computer Science · First Class

2021 — 2025

05 / What's next

Let's build something worth understanding.

Exploring opportunities in software engineering, applied AI and research engineering.

Download CVOpen GitHub profile中文简历

Public CVs omit personal contact details. Connect through my GitHub profile.