SolutionStreet

Contact us for a Free Consult Now!

We are AI & ML Experts

With more than 15 years of working with AI and machine learning, Solution Street’s engineering and delivery teams are experienced with advising, implementing and operating iterative AI & ML solutions–from NLP and OCR to predictive analytics.

How we help organizations overcome internal barriers to get the most out of AI

Our process starts by working with clients to diagnose their unique business challenges. We then identify which problems are best suited for AI solutions and prioritize them based on potential business impact. Our team then dives deep into the data, uncovering insights and patterns that help inform our approach. Through collaboration, we refine our focus and develop a clear understanding of what implementations will be the most important and impactful. From there, we clean, develop, validate, and rigorously test the data before deploying a tailored AI model designed to drive measurable growth for your company.

Our engineers, architects and developers here at Solution Street have significant experience working with and building systems with Artificial Intelligence and Machine Learning, and are ready to help you get the most out of AI.

 

Solution Street ML Process

Solution Street ML process

Contact us for a Free Consult Now!

Service Delivery Experience

Solution Street has successfully delivered many cutting-edge AI and ML enabled platforms over the past 15 years. Unlike some other firms that are drawn to the new allure of AI, our business prides itself on our experience in the subject and the deliverables we've developed! Below are some examples of projects we have worked on:

Machine Learning Regression-Prediction

Business Situation: Solution Street’s customer in the shipping logistics industry can incur costly backups when unforeseen weather delays hit their key ports.

Technical Situation: We leveraged Azure AutoML to build machine learning models including Logistic Regression and Random Forest trees to predict shipping delays.

Solution: After we constructed a predictive model based on historic weather delays and related weather conditions, we built a system to feed weather forecast data into our model in near real time along with current shipping lineups to predict possible weather delays, thereby allowing our customer to make adjustments proactively accounting for significant savings in delay costs.

Entity Recognition, Semantic Similarity

Business Situation: Solution Street’s customer in the music royalty clearing space had a situation where authors, artists, song titles, and versions were being imported in a non structured format making it difficult to determine matches.

Technical Situation: We leveraged AWS Sagemaker & BERT based models to construct machine learning models to identify and classify entities.

Solution: We created network embeddings of the "authors collaboration network" (e.g., author A works with author B on a piece, and author A works with author C on another piece) to find "similar" authors. We can then correct incomplete names with confidence (e.g., the closest author to J. Lennon and Paul McCartney is "John Lennon" (in terms of closest in the network), so we can say "this J. Lennon is actually John Lennon" (there might be an author called James Lennon, but that one is not as closely related network wise).

Natural Language Processing

Business Situation: Solution Street’s customer wanted to cluster text documents at scale, in this case it was clustering websites.

Technical Situation: Leveraged Hadoop clustering and open source NLP algorithms to accomplish clustering.

Solution: Documents were analyzed using Term Frequency - Inverse Document Frequency (TF-IDF) to generate vector representation of the documents. Leveraged K means clustering algorithm over the the document vectors. Principal component analysis was experimented with to reduce the dimensionality of the data. In the end, millions of websites were clustered into groupings such as finance, banking, sports, news, weather etc.

Retrieval Augmented Generation

Business Situation: Solution Street has posted over 150 blog articles that share the experience of its key engineers over a course of 20+ years. We wanted the ability to share similar articles from their library with their readers.

Technical Situation: Leveraged Retrieval Augmented Generation to create an extension of a large language model to show similar blog postings.

Solution: We constructed an embeddings database of all of our blog articles and added it to an existing LLM to achieve the desired result.

"Solution Street's use of Azure AutoML and machine learning models has transformed our approach to managing shipping logistics. Their predictive system uses real-time weather data to foresee and manage potential delays at our key ports, significantly reducing delays which leads to overall cost savings for us."

Molly, Managing Director of Logistics

Let us help you with your next Artificial Intelligence project. Contact us today!