Avatar

李大為

量化策略機器學習研究員

靖戈

自我簡介

我是李大為 (David Lee),是前微軟資料與應用科學家,目前在一家私募量化裡做機器學習量化策略研究。自許為 maker —— 努力實現任何想到的有趣點子。我喜歡音樂和其他我覺得很酷的東西。有輕微的強迫症,做事必須要有條理,特別是在coding上必須規範且優雅。秉持能隨便就隨便但要拘謹就該嚴格的人生哲學(自稱為「真值表哲學」),簡化太過複雜的人生,努力為了變得更懶惰而勤勞。學習上,求知若渴,虛心若愚,也不吝於與他人分享自己的所學,正所謂開源精神。

興趣

  • 人工智慧
  • 自然語言處理
  • 量化交易
  • 推薦系統
  • 嵌入式系統設計

教育程度

  • 軟件工程 工學碩士, 2021

    北京大學

  • 電子工程 工學學士 (輔修財金), 2017

    國立臺灣科技大學

專長與興趣

當一個人對一件事足夠感興趣,那麼他將會努力讓它成為自己的專長之一。

Programming

Python 機器學習
C/C++ 嵌入式系統設計
Node.js 後端設計
Java Android App 設計
C# Unity 遊戲設計
Verilog HDL FPGA 晶片設計
Matlab, R 數學統計計算

熟練使用 Vim

Music

爵士鼓
吉他
鋼琴
管樂團打擊樂
宅錄

Other Hobbies

攝影
咖啡(拉花、手沖)
溜滑板
魔術方塊
3D 列印
直排輪
騎腳踏車
滑雪
騎摩托車
飛無人機

刊物

Open Relation Extraction via Query-Based Span Prediction

QORE utilizes a Transformers-based language model to derive a representation of the interaction between arguments and context, and can …

Open Relation Extraction with Non-Existent and Multi-Span Relationships

We proposed a Query-based Multi-head Open Relation Extractor (QuORE) to extract single/multi-span relations and detect non-existent …

Towards Topic-Aware Slide Generation For Academic Papers With Unsupervised Mutual Learning

Generating slides from papers by extractive summarization techniques and unsupervised mutual learning to deal with data lacking issue.

生活

一生只活一次,所以放手一搏吧!

.js-id-Music

微軟蘇州樂團 - Return True

我們因為 FY22 Kickoff (2021年8月) 的表演而相聚,並一起玩音樂直到今日。我在樂團中擔任鼓手、木吉他手、鍵盤手、抒情人聲、錄音師、混音師、…

經驗

工作 / 實習

 
 
 
 
 

量化策略機器學習研究員

上海靖戈智能科技有限公司

May 2023 – 現在 上海
程序化T0交易量化機器學習策略研究
 
 
 
 
 

資料與應用科學家

微軟亞洲互聯網工程院 - WebXT Bing 多媒體

Jul 2020 – Mar 2023 蘇州
在Bing、MSN、Edge上的相關影片推薦系統
 
 
 
 
 

算法實習生

微軟亞洲互聯網工程院 - WebXT Bing 自然語言處理 Carina

Jul 2020 – Jun 2021 北京
Currently working on Writing Assistant related projects with NLG techniques. Including data collection, model training, and backend API services hosting.
 
 
 
 
 

研究實習生

微軟亞洲研究院 - Knowledge Computing

Dec 2019 – May 2020 北京
Take over mainly two research-oriented NLP projects.

  • Generation of slides from academic paper
  • Math word problem generation
 
 
 
 
 

研究實習生 (實驗室)

北京大學 軟件工程國家工程研究中心

Jul 2019 – Jun 2021 北京

Doing case of Anti-healthcare fraud and Medical record analysis.

Including research of:

  • Information Extraction
  • Named-entity Recognition
  • Relation Classification
  • Knowledge Graph

北大學位論文: 面向中文文本的數值事實抽取方法設計與實現

 
 
 
 
 

嵌入式系統軟體組實習生

工業技術研究院

Jul 2016 – Aug 2016 新竹
I was in the self-driving group, I mainly handled the STV0991 development board which was going to carry the computer vision algorithms.

個人接案

 
 
 
 
 

腦電圖濾波與視覺化分析

台科大企管系教授

Sep 2018 – May 2019 遠端
I used Matlab to process and analysis EEG raw data. And do some visualization and animation on it.
 
 
 
 
 

Leapsy AR 眼鏡影像串流遙控攝像雲台

All Joint 傲嬌文創

Jul 2017 – Oct 2017 台北
I collected sensor data on Android-based AR Glasses to capture current attitude and sent it back to Raspberry Pi to synchronize camera pan/tilt head’s direction then return video stream back to glasses through Wi-Fi. And I made pan/tilt head structure using 3D print model to contain camera and two servo motors, and designed the power supply circuit for both motors and Raspberry Pi.
 
 
 
 
 

心電圖分析

NTU on-the-job Ph.D. Student

Nov 2015 – Feb 2016 遠端
I used Matlab to do fourier transformation on ECG (Electrocardiogram) signal by filtering out the high frequency noise and finally predicting its trend.
 
 
 
 
 

油品監測系統

貝爾特

Jul 2014 – Oct 2014 遠端
I and my collage roommate Tom built a Windows application to get the machine’s sensor values, show them and store them in a database. This project was asked to use Visual Basic.

比賽

 
 
 
 
 

BeChangeMaker

World Skills

Mar 2023 – Sep 2023 線上

Ecojoy

We want to solve the problem of “Toy waste”. Excessive pollution not only affects the physical environment of future generations but also cultivates children who do not cherish resources, which has a major impact on the world. We hope that through a very simple way, every old toy will no longer be piled up at home or enter the landfill, but can also become a resource for others. We have software engineering, social education, and economics background. Observing that the problem of toy waste is becoming more and more serious, it is readily available and cheap, becoming a quick solution for most parents to deal with their children. We believe that as long as the sharing and acquisition methods are simple enough, it can immediately improve the situation of excessive waste. Through subscription to become members of Ecojoy App, you can easily share excess toys at home, and through the perfect toy information and rating system on APP, users can easily find suitable toys to meet their needs and achieve toy sharing and reuse.

Facebook Page

 
 
 
 
 

Jigsaw Unintended Bias in Toxicity Classification

Kaggle

Feb 2019 – May 2019 線上
This competition is aim to classify whether a comments is toxic. Our team design different models such as BERT, ELMo etc. as classifier and finally ensemble them. Our team reach Top 1% in rank.
 
 
 
 
 

混凝土泵車砼活塞故障預警

Digital China Innovation Contest 2019

Jan 2019 – Mar 2019 線上

In this competition, each sample is a time-series data of a concrete pump vehicle. The goal is to predict the likelihood of each data sequence that whether a machine might fail. I used LightGBM and reach Top 5% in rank.

Source Code

 
 
 
 
 

ARM Design Contest

ARM

Apr 2016 – Nov 2016 新竹
Based on my independent study of department project - the quadcopter project. Using specified development board STM32F4 to drive the quadcopter. We get Top 10 in the final.
 
 
 
 
 

盛群盃 MCU 設計比賽

HOLTEK

Apr 2016 – Nov 2016 台中
Based on my independent study of department project - the quadcopter project. Using specified development board STM32F4 to drive the quadcopter. Finally, we get honorable award.
 
 
 
 
 

台大三校聯盟 App 創意競賽

台灣大學聯盟(台大、台科大、台師大)

May 2015 – Aug 2015 台北
Designed a platform called Skill Exchange - a talent and skill exchange platform which matches people with their know-how and what they want to learn. Finally, we get honorable award.
 
 
 
 
 

中山大學 LED 創意設計競賽

中山大學電機系

Oct 2014 – May 2015 高雄
An installation art LED grid ball that combined sound and light. This project collaborated with design department students. Using gaming button to trigger MIDI signal to a computer to make a sound. And control LED grid with Arduino. Finally we get merit award.
 
 
 
 
 

台灣大學 Taiwan 2048 BOT 大賽

台灣大學

May 2014 – Jul 2014 台北
I and my friend Tom built an AI BOT for the 2048 game. We used Monte Carlo Tree Search (MCTS) with alpha-beta pruning to select best action. And score each state(board) with our own designed evaluation function. Finally we get honorable award.

證照

中級咖啡師

江蘇省職業技能證書,編號 S000032050806234001680
查看證書

TOEIC(多益) 785990

Test of English for International Communication: Advanced
查看證書

電腦硬體裝修 乙級

查看證書

電腦硬體裝修 丙級

查看證書

項目

課外項目 / 課程項目 / 原始碼

*

Stanford CS224n NLP with DL

自學課程,其中包括 word2vec、dependency parsing、machine translation、question answering 等 projects。

SemEval-2013 Word Sense Induction

SemEval-2013 Task 13 Word Sense Induction for Graded and Non-Graded Senses.

Jigsaw Unintended Bias in Toxicity Classification

Kaggle 的比賽,目標是要判斷一些網路上的評論是否為 toxic。

SemEval-2018 Relation Classification

SemEval-2018 Task 7 Semantic Relation Extraction and Classification in Scientific Papers.

混凝土泵車砼活塞故障預警

基於特徵工程的比賽,對於 time-series 的運轉數據預測其故障的可能性。

Operating System

PKU OS course project and notes based on Nachos and XV6

2048 AI BOT

搭建 2048 AI BOT。在 2014 年的比賽中搭建 MCTS 版本,並於 2018 年的 AI 課程中搭建強化學習版本。

Raspberry Pi Cluster GitHub stars

開源的 quick-start 工具,可以快速搭建 Raspberry Pi Cluster 並搭載一些著名的 ecosystem 例如 Hadoop, Spark 等。

Deep Learning Practice GitHub stars

實作神經網路,其中包含各種如 NLP、RL、CV 相關 topcis 的項目。

Machine Learning Practice GitHub stars

從零實作機器學習演算法,其中包含許多課程項目與筆記。

基於模組化架構之四軸飛行器設計及其於影像辨識之應用

大學畢業專題。從零打造四軸直昇機,並且在多種不同的開發平台上運行,並結合電腦影像技術做自動控制。獲得校內最佳專題獎、最佳人氣獎。

領導能力與課外活動

.js-id-Leadership

台科電子系學會

Serve as atristic designer. Handle poster design and Facebook fans page operation.

二重國中管樂校友團

Being principal percussionist in junior high school alumni wind band during 2016, 2017, 2018, 2019 summer.

台科竹友會

Serve as photographer and social media manager. Serving in hometown for elementary school students in 2014.

腳踏車環島

Cycling counter-clockwise around Taiwan with senior high school classmate in ten days.