AI Copilot for Industrial Process Control

Shuhan Technologies Co., Ltd. strives to provide AI solutions to process-control in industrial manufacturing. Our flagship product is an AI-predictive maintenance system (an autonomous control system) for industrial plants, with its implementation focusing on increasing safety and economic value of plant operations.After a series of successful large-scale projects in China, our company now expands globally.Our current research focuses on industrial applications of Large Language Model (LLM), specifically revolving around the AI-Agent paradigm.

My role:- Managing the R&D and applications of AI technologies in the field of environmental protection, particularly in wastewater treatment.Technologies developed by the company:- autonomous control system for wastewater treatment units- FT-NIR MEMS chipset- Autonomous spectral analysis system, based on Near-Infrared Spectrocopy, using Machine Learning to algorithmically decompose Adsorption Spectrum, predicting chemical composition of wastewater.- real-time (computer vision) video detection of turbulent flow, and control strategies- wastewater load prediction system

Currently managing the tech development of an AI startup company, a spinoff from 浙江巨能环境工程.The AI startup develops a Predictive Maintenance System for Wastewater Treatment Plants, empowered by Reinforcement Learning and IoT, to optimize the operations and controls of wastewater treatment processes. Currently our AI system is under trial-phase in some large industrial wastewater treatment plants in China.Fields/Disciplines involved:Machine Learning, IoT, Web Development, MEMS Chip Design, Signal Processing

My role is to lead a small team of AI/ML researchers and web developers to develop an AI that automates client research, engineering process design, budgeting and 3D modelling, in the field of wastewater treatment engineering. My current work involves:-Evolutionary Strategies-Reinforcement Learning, Q-learning-Unsupervised Learning Methods, such as Autoencoder-decoder, Generative Adversarial Network, Restricted Boltzmann Machine, kNN-Regression-Meta-Learning, efficient learning with small datasets-Time-Series Analysis, Forecasting, Signal Processing for system of sensors-Anomaly detection (Past) concepts/framework that I frequently dealt with:-Web scraping -Mandarin Natural Language Processing to extract Entities and Relationships, to form RDF triples-Knowledge Graph, Neo4j-Knowledge Graph-based Q&A system-Graph Algorithms, Graph-based Logical Inference-Graph-based Machine Learning, especially unsupervised method to classify node labels etc.-Web UI development for engineers' use-Image Recognition: extracting information from graphs, process-flow diagrams and tables that contain uncommon technical terms.

My work frequently uses and deals with the following concepts/framework:- Supervised Learning: Penalized Linear Regression (Lasso, Ridge, Elastic), LARS, Kernel Trick, Glmnet, K-Nearest Neighbor, Bayes Classifier, Forward Stepwise Regression, Support Vector Regression- Unsupervised Learning: K-means clustering- Reinforcement Learning- Decision Trees, Bagging, Random Forest- Support Vector Machine with Kernels- Genetic Algorithm- Multilayer Perceptron, Deep Neural Network using TensorFlow with GPUs- Scikit-learn- Machine Vision, OpenCV, Image Processing, YOLO- First-order Logic Inference & Knowledge Graph: Markov Logic Network, Random Path Algorithm- Convolutional Neural Network for image recognition, Recurrent Neural Network for sequential data such as speech and texts (Natural Language Processing).- Distributed Computing & Data Manipulation: Pandas, Apache Spark, Hadoop Cluster Computing- Web scraping: BeautifulSoup & PyQuery, Pyspider (distributed web-scraping framework), Chrome Puppeteer, Selenium- Kafka, Redis, MongoDB, HDFS, MySQL- Perceptual Hashing Algorithms- Deployment: Docker, Flask- Visualisation: Matplotlib, tk, PyQt, Kibana, Tensorboard- Search Algorithm: Elasticsearch, Vector Space Model with KNN- Language Processing (Mandarin): HanLP, jieba

Help Yudu start up and lead a project on cross-border ecommerce. First three month the team was doing drop-shipping and private-labeling operations on AliExpress, Amazon and Lazada. Our primary consumer locations are in the US, Russia and South East Asia. - Negotiated with suppliers of chinese-made products of high quality, and sold them on AliExpress, Amazon and Lazada.- Advertized and marketed on Facebook and through Google SEO, brand fan page grew from 0 to 35000+ followers in just 7 days.- wechat marketing using subscription accounts, to attract aspiring entrepreneurs, and we teach and train them on doing drop-shipping and private-labeling on cross-border ecommerce platforms.For the first three months, we were burning cash on marketing and loss-making, but starting from October, we started to become profitable, and in the November, due to Black Friday/Thanksgiving Day/Cyber Monday, we generated a net profit of 30K USD, pretty good for starters like us in our first year doing this.

Based in Wuzhen where the Wuzhen World Internet Conference is held every year, I help Yudu Group's CEO and Yudu's affiliated VCs analyze tech investments and projects of Yudu in China. Projects and investments that I analyzed: - Electric vehicle timesharing app （新能源汽车分时租赁）- C2C Uber-style logistics app- Cross-border ecommerce, analyzing supply, marketing and operations of Alibaba, AliExpress, Amazon, Wish, Ebay and Walmart. （跨境电商运营）- Taobao, Tmall and JD operations.- Cross-border ecommerce entrepreneurship training, on developing a franchise model.- Supply-chain financing credit analysis, dealing with 300 Million USD micro business loans, using big data（供应链金融）- Wuzhen Institute's Knowledge Graph （乌镇智库知识图谱）businesses, on using discrete mathematics to generate a knowledge graph with a certain algebraic structure, which scrapes and crawls the entire web and applies set theory to evolve itself. Because of its machine learning techniques, the knowledge graph has predictive power, and can generate highly industrial specific reports. Reports are usually sold to Alibaba, Tencent, China Mobile, Research Institutes and other big organizations, one report can be sold up to 1 million USD.- IoT, on Wifi probing technology using a modified wifi router（WiFi探针物联网技术）, and gathering data to construct foot traffic data and trajectories of people nearby the wifi router. These data can then be used to trade for user-specific data, and sold to analytics/marketing companies. Particularly useful for offline retail and logistics operations.- O2O and B2B operations of Clothing Supply Chain Platform (Haoxiadan.cn, 好下单平台，3.5 Million USD Series-A funding in 2016). Particularly, on how to build a platform startup in the mobile ecommerce space.- Blockchain- Using anaerobic bacteria to digest organic waste in the wastewater treatment system, with a low carbon footprint.- B2B offline event and popup space renting platform

Over the past 5 years, I've been coaching on areas such as business, psychology, math, Oxbridge/Ivy League admissions, web development, learning, academic success, personal development, language learning, piano playing, meditation, etc.My passion is to inspire people to chase their dreams. I empower people's psychology to help them achieve their success.

HireSolve is a recruitment platform that connects recruiters and job applicants through problem-solving. Recruiters can post problems relevant to jobs/roles on the platform. Applicants can apply to jobs by submitting solutions (in text, image, video, code snippets, etc) to the posted problems. If a recruiter likes a solution, the applicant behind the solution will get contacted for further interview. HireSolve is founded on a belief that, if you can solve a problem, you demonstrate that you have the necessary skills. This is much more effective than resumes, because resumes can be faked and thus not reliable. By focusing on problem-solving, we can reduce bad hires, save money and time for businesses, and we cut out the need to screen resumes.We built the demo for HireSolve in the AngelHack KL 2016, and we're selected as a finalist from 84 competing teams.We plan to scale HireSolve in both SEA and India at the same time. We believe that HireSolve can impact the entire recruitment industry, and make the hiring more effective and efficient.We've validated the business model in India (and earned some revenue out of it!), and we're running some validation cases and receiving pre-orders in Malaysia and India. We're now funded by an angel investor, to support our backend development, which we shall launch publicly on 1st September 2016. We've launched a 2000 USD online challenge, for anyone to contribute us great ideas that can help us go global in the early stage. We'll reward 2000 USD for any idea that resonates with us, no questions asked, no contract, no terms & conditions. Participate in the online challenge here: http://crazyidea.hiresolve.co/We're also looking to hire a full-time marketing talent.

JomSpace connects people and businesses through space-sharing. JomSpace strives to make space-sharing fast, easy and social.Our vision is to become a leading global platform for space-sharing, and pop-up retail. We want to help reduce the barriers to off-line retail, and we believe pop-up retail is the future of retail. We're also helping entrepreneurs, by making off-line retail more accessible and less costly. We treat our business as a way to learn about the fundamentals of building a platform business.Ultimately, we want to empower humanity at a great scale.JomSpace is now a team of 4 energetic and visionary young graduates (two of us are from Cambridge University!), running marketing and website development.

I see teaching as a great way to train ourselves some entrepreneurial skills, such as having the growth mindset, leadership, communication, marketing and coaching skills. Mostly for me, it's about overcoming the fear of dealing with people, and to learn how to motivate others to pursue their dreams.Through my teaching career, I learn to appreciate a lot of things that I took for granted. I also learned that everyone has the seed of greatness, and the reason to be great. Everyone wants to be great. Education is a pathway to greatness.Most importantly, teaching is a respectable and noble job. It inspires young people to chase their dreams, and it opens up our mind.