Let’s build together! I provide consulting services in ML/AI Engineering, Full-stack Development and Engineering Management.

I can help you with both design and implementation of your solutions. I am also experienced in training teams (see conferences).

My long-term engagements and experience can be found on my LinkedIn profile.

You can book a paid session below or reach out via email: office@horosin.com.

Some of the areas I can help with:

• Design and development of generative AI systems – implementing AI products based on LLMs, coordinating AI projects, MVPs and production systems. I have experience with Langchain (plus Langserve, Langsmith, LangGraph). I teach people in this area on my blog.
• Software engineering - I am an experienced engineer and I delivered many successful projects in my career. I am able to handle variety of technologies and my expertise lies in Node.js, Python, Frontend (React/Vue/others), data engineering, large-scale serverless cloud computing.
• Writing grant applications – architecture of AI projects, designing research methodologies, and outlining initiative significance and potential challenges.
• Training – teaching practical skills in AI Engineering, ML Ops, Node.js, JavaScript, Python, full-stack development and cloud computing . Also trained teams in agile processes, system design and architecture (see my public appearances). Reach out if you need a workshop or training.
• Architecture consulting – developing long-term strategies aligned with business objectives, diagnosing and resolving technical issues.

Case studies

Two of my most recent projects.

AI Knowledge Discovery System for AEC (Architecture, Engineering and Construction)

Client: An AI-powered knowledge discovery startup in the AEC industry

Technology: Langchain, Python, Langserve, PGVector, PostgreSQL, Supabase, Deno, Node, OpenAI API

This project aimed to implement knowledge discovery backend for Architecture, Engineering, and Construction (AEC) professionals by integrating a Retrieval-Augmented Generation (RAG) system. The RAG model leveraged advanced retrieval and generative AI to deliver precise, context-aware responses, enabling novel optimisations in AEC processes.

We implemented an AI agent capable of using multiple tools (SQL, Vectorstore and others). The system was self-correcting to provide structured and useful responses. We developed an ingestion pipeline for diverse AEC datasets using PGVector with OpenAI embeddings and PostgreSQL for structured data. We used Langchain to build the OpenAI-based agent, and deployed the system using Langserve, Deno, and Node. The database and processing was implemented in the Supabase ecosystem.

The project delivered our client an essential part of their product and aided in their fundraising efforts.

Self-Hosted Fine-Tuned Coding Assistant

Client: An enterprise with 5,000+ employees

Technology: Custom AI models, Internal IDE plugins, Deployment tools

This project involved designing and planning the development of a bespoke coding assistant tailored for a large enterprise, necessitated by the prohibitive costs of existing solutions like GitHub Copilot, which approximated $2 million annually. Additional motivations included specific contractual obligations with customers and the need for a system optimized for the unique performance requirements of the company’s industry.

The development strategy encompassed fine-tuning an AI model on proprietary code, creating custom IDE plugins, and establishing a comprehensive testing and rollout plan. This in-house solution was presented to the company's board of directors, highlighting its potential to significantly reduce costs and enhance coding efficiency for developers.

Testimonials

Here are some people that liked working with me.

Book a consulting session

Schedule an ad-hoc session with me by using the provided form. If you need a quote for a larger initiative, reach out via email office@horosin.com. Please provide as much detail as possible regarding your inquiry.