Founder | Machine Learning Engineer
2020 - PRESENTMLAI, LLC.- Conceived, designed, and built an end-to-end web app.
- Automated the daily refresh of news content enabling the entire app to work on autopilot.
- Manually labeled data to train a model for clickbait detection in news articles with 83% accuracy.
- Optimized the performance to efficiently fetch around 5,000 distinct news articles every day from 400 news feeds.
- Fetched and parsed news articles daily, updated models and predictions and uploaded an optimized view to AWS S3.
Technologies: Natural Language Processing (NLP), Python 3, Flask, JavaScript, JSX, SpaCy, Bootstrap, RSS Feeds, JSON REST APIs, SaaS, Machine Learning, Exploratory Data Analysis, Pandas, Amazon Web Services (AWS), RESTful Development, SQL, Python, Statistical Learning, Deep Learning, NumPy, Data Science, REST, Fast.ai, React, Full-stack, Visual Studio Code, PostgreSQL, Seaborn, Keras, Deep Neural Networks, Algorithms, MacOS, Heroku, APIs, Architecture, Artificial Neural Networks (ANN), Sentiment Analysis, Matplotlib, Scikit-learn, SciPy, Systems, Neural Networks, NLTK, Gensim, Semantic Analysis, Tf-idf, REST APIs, Web Scraping, Task AnalysisAI/ML Engineer
2022 - 2022Healthcare Provider Client (via Toptal)- Conducted feasibility analysis of an ML model for the business problem.
- Created estimates of data collection and annotation needed.
- Provided candidate models to evaluate once data was available.
- Answered all client's questions to their total satisfaction.
Technologies: Machine Learning, Deep Learning, Computer Vision, Healthcare, 3D Image Processing, Image Processing, Medical Imaging, Task AnalysisSenior AI Engineer
2021 - 2022RedRoute- Owned AI for the company; defined the AI roadmap for the company, combining several ideas from business, product, and technology as well as from academia, research, and industry.
- Led the audio-related data science roadmap and work; mentored/technically managed an experienced audio data science researcher.
- Built a real time audio barge-in detection model that performed very well with low resource requirements, compared to several prior attempts.
- Improved the performance of an eCommerce intent detection model by 2-5%—analyzing and relabeling the data, manually prioritizing and labeling high impact utterances, retraining and redeploying the model, and creating and monitoring looker dashboards.
- Increased customer handle rates by adding richer responses for eCommerce customers; this was done by answering frequently asked questions based on information from the customer's website and knowledge base. Created dashboards to monitor these changes.
- Created a transcriber/speaker diary tool to transcribe conversations.
- Built several dashboards and looks (queries) in Looker to monitor the impact of different changes.
- Educated the team on Kanban and influenced the deployment of a variant of Kanban in the company.
- Contributed to other areas of the company, including preparing to scale and establishing and improving processes and the business/sales/marketing/product strategy.
Technologies: Python, Python 3, TensorFlow, Deep Learning, Realtime, IVR, Interactive Voice Response (IVR), Dialog Systems, Amazon Web Services (AWS), Ansible, Docker, Machine Learning Operations (MLOps), Google Speech-to-Text API, Labeling, Jupyter Notebook, Jupyter, Kanban, Looker, ETL, MongoDB, Snowflake, Flask, Technical Hiring, Source Code Review, Code Review, Task Analysis, Interviewing, Team ManagementData Scientist
2018 - 2020Microsoft- Built models to evaluate over 100 datasets to discover methods to improve the AutoML library.
- Improved the product on the benchmark against the competition by investigating and explaining a competing AutoML product's scoring methodology for imbalanced multi-class classification. I did this by digging deep into how metrics were computed.
- Analyzed feature importances by training over 100 datasets with automatic feature engineering libraries to prioritize featurizers for our AutoML product.
- Developed an end-to-end AutoML framework as part of a Hackathon project to understand what approaches would work best for AutoML.
- Delivered a dataset analysis and onboarding tool, which enabled evaluating, filtering, cleaning, and onboarding of 20 datasets into our benchmarking corpus.
- Built better complete performance graphs to improve confidence in the performance of each competitor across a corpus of over 100 datasets.
- Identified gaps in our benchmarking dataset corpus distribution and added over 20 datasets to fill those gaps.
Technologies: R, Microsoft Power BI, Matplotlib, Scikit-learn, AutoML, Azure, Jupyter, Pandas, NumPy, Python, JSON REST APIs, REST APIs, SaaS, Machine Learning, Exploratory Data Analysis, Python 3, RESTful Development, Docker, SQL, Statistical Learning, Deep Learning, Data Science, Visual Studio Code, Artificial Intelligence (AI), Seaborn, Azure IaaS, TensorFlow, Keras, Algorithms, Linux, Windows, APIs, Architecture, Artificial Neural Networks (ANN), SciPy, Tidyverse, Ggplot2, Systems, Neural Networks, Azure Machine Learning, Flask, Agile, Source Code Review, Code Review, Task AnalysisFounder
2015 - 2019Meon- Built a web platform that allows you to create apps in minutes, Meonapp.com.
- Developed 80 apps in two days across several application domains.
- Created apps for various clients on the platform, including a tailor management app, a music composition app.
- Added capabilities for both general users and developers to build apps.
- Enabled apps to be hosted as soon as built, reducing turnaround time greatly.
Technologies: Heroku, PostgreSQL, Ruby on Rails (RoR), Ruby, SaaS, RESTful Development, jQuery, JavaScript, SQL, REST, Full-stack, Visual Studio Code, Algorithms, Linux, MacOS, Java, APIs, Architecture, Systems, REST APIs, Source Code Review, Code Review, Task AnalysisSenior Software Engineer
2017 - 2018Divensi, Inc.- Researched and developed a deep learning model for 3D point cloud semantic segmentation of imbalanced outdoor Lidar data, with near state-of-the-art results for outdoor Lidar.
- Developed a V1 cloud-hosted decision support system web application utilized by several enterprise users for a remote startup client. Hired a technical team and transitioned the product. Offered a CTO position by the client CEO.
- Built a data pipeline framework for machine learning experimentation with Lidar data.
Technologies: PDAL, Laspy, LiDAR, Jupyter, TensorFlow, Python, JSON REST APIs, Machine Learning, Exploratory Data Analysis, Python 3, Pandas, RESTful Development, JavaScript, Google Cloud Platform (GCP), SQL, Deep Learning, Statistical Learning, NumPy, Data Science, REST, Fast.ai, Full-stack, PostgreSQL, Artificial Intelligence (AI), Computer Vision, Keras, Django, Django REST Framework, Deep Neural Networks, Algorithms, MacOS, APIs, Architecture, Artificial Neural Networks (ANN), Image Recognition, Matplotlib, Scikit-learn, Convolutional Neural Networks, SciPy, Systems, Neural Networks, Agile, REST APIs, 3D Image Processing, Image Processing, Technical Hiring, Source Code Review, Code Review, Task Analysis, Interviewing, Team ManagementFreelance Developer
2012 - 2014Self Employment- Built nine educational games that were released to the App Store; iOS and cross-platform using the Corona SDK.
- Performed App Store optimization to maximize adoption and saw over 30,000 downloads across games.
- Developed a broad range of games, from running quizzes to new mathematical puzzles. The samurai game was highly appreciated by middle school teachers in the US.
Technologies: ASP.NET, Corona SDK, Objective-C, REST APIs, SQL, iOS, Algorithms, MacOS, Architecture, Systems, Task AnalysisFounder
2009 - 2011Thouwords, LLC.- Built a web application to make a textual website more visual by performing topic modeling with Alchemy API. Obtained images for each topic from Wikipedia APIs.
- Created a Wikipedia visual navigator using topic modeling and Wikipedia API to get images.
- Developed a web application to create rich ebooks for kids using pictures, video, and text books. The application was used to create picture books for kids and shared with parents.
Technologies: AlchemyAPI, ASP.NET, Machine Learning, JavaScript, SQL, Statistical Learning, Algorithms, Service-oriented Architecture (SOA), APIs, Architecture, Sentiment Analysis, Systems, Semantic Analysis, REST APIs, Technical Hiring, Source Code Review, Code Review, Task Analysis, Interviewing, Team ManagementSenior Technical Program Manager
2000 - 2009Microsoft- Built BizTalk Server, an enterprise messaging and workflow platform from idea to product. Released three versions of BizTalk Server.
- Created the first .NET based Outlook API, and released two versions of it.
- Built a service delivery platform for mobile telecom providers using .NET and SOA, including several WS- standards like WS-reliability and WS-eventing.
- Defined and led the inclusion of the REST API in .NET WCF.
Technologies: Windows Communication Framework (WCF), Service-oriented Architecture (SOA), Outlook, BizTalk Server, .NET, JSON REST APIs, SDKs, SQL, REST, C++, Algorithms, Windows, ASP.NET, C#, APIs, Architecture, Systems, Workflow, RSS Feeds, Agile, REST APIs, ETL, Technical Hiring, Task Analysis, Interviewing, Team ManagementTechnical Product and Program Manager
2006 - 2007Amazon- Captured the end-to-end Amazon retail messaging and workflow blueprint by collaborating with over 40 teams at Amazon that used the messaging and workflow framework.
- Defined and led the creation of an internal distributed configuration store modeled on DNS.
- Proposed a well-received vision for (the then new) Amazon cloud. It was based on true elasticity and automatic scalability.
- Acting upon the vice president's suggestion, I presented the proposal to a special future architecture group. They incorporated it into their plans.
- Drove the adoption of a distributed configuration store across teams in the company.
Technologies: Java, Workflow, Engineering, SDKs, SaaS, Amazon Web Services (AWS), REST APIs, RESTful Development, SQL, C++, Algorithms, APIs, Architecture, Technical Hiring, Task Analysis, InterviewingSenior Software Engineer
1998 - 2000Microsoft (via Aditi)- Built a code profiler for Visual Studio Internal tools in C++.
- Developed an XSD (XML Schema Definition) library in C++.
- Created a persistence layer for an XSLT based BizTalk schema mapper.
- Ported the FrontPage server extensions to a pre-release version of C#.
Technologies: Profiling, CODE, Visual Studio, XSLT, XSD, XML, ATL, C++, SQL, Algorithms, Windows, ASP.NET, .NET, C#, APIs, Architecture, Systems, Technical Hiring, Source Code Review, Code Review, Task Analysis, Interviewing, Team Management