Adam Sroka

Data & AI Director

Hypercube Consulting

Chief Executive Officer

Director

The Data Lab - Innovation Centre

The Innovation Advisory Board is responsible for advising on innovation and collaborative project activity for The Data Lab – Scotland’s Innovation Centre for Data and AI.

Innovation Advisory Board Member

Origami

Head of Machine Learning Engineering

Incremental Group

Senior Data Scientist

Sure Thing! Insurance

Data Scientist

Clear Returns

Thales

Research Engineer

I build data and AI platforms for the energy sector | Director at Hypercube Consulting

Dr Adam Sroka is a Co-founder and CEO at Hypercube Consulting, a firm that helps organisations unlock value from data by delivering enterprise-scale solutions and building high-performing data and analytics teams from the ground up. With over 10 years of experience in data science, machine learning, and artificial intelligence, Adam is a recognised expert and speaker in his field.

Adam's mission is to bridge the gap between technology and business, and to enable data-driven innovation and transformation across various industries. He shares his insights and best practices through his blog, podcast, and tech community events, where he showcases his approaches and systems for effective communication, engineering discipline, and experimentation. Adam is passionate about helping organisations and data professionals maximise their potential and impact with data and AI.

iCustoms

- Designed and deployed an AWS-based, event-driven Intelligent Document Processing (IDP) system using MLOps practices, integrating autonomous agent workflows, real-time inference, and RESTfulAPIs (FastAPI) to automate back-office operations, reducing document processing time by 70%. 
- Developed a multi-agent LLM system with LangGraph and OpenAI(GPT-4omini) for HS code classification, implementing RAG with vector databases (ChromaDB) for context-aware, dynamic decision-making across 5+ countries.
- Built a web app for human-in-the-loop data annotation, enhancing efficiency by 30% and enabling scalable dataset preparation for transformer-based models. 
- Fine-tuned a multimodal LLM using LoRA for Named Entity Recognition (NER), achieving 95% extraction accuracy for customs documents, with prompt engineering frameworks to ensure consistent outputs. 
- Integrated Redis for in-memory caching in an AI-powered HS code classification system with asynchronous API design, cutting inference latency by 40% and ensuring seamless integration into production systems with safety guardrails. 
- Established a benchmarking pipeline with robust evaluation methodologies to monitor and optimize ML/LLM performance, scalability, and safety for IDP, incorporating R&D on SOTA LLMs and classical ML techniques. 
- Developed and deployed a conversational AI agent using OpenAI to handle customer queries about iCustoms, enhancing user experience and reducing query resolution time in a scalable production environment. 
- Built a search system for HS code lookup using keyword search and vector stores, enabling users to input product-related keywords and receive real-time suggestions of 10-digit HS code descriptions in a dropdown for multiple countries.

- Managed data collection and labeling of a 100K+ product description dataset, collaborating with a cross-functional team of 5 to support text classification models, incorporating feature engineering and EDA.
- Implemented automated data preprocessing pipelines using Python (Pandas, NLTK) to enhance efficiency, reducing dataset preparation time by 50% for downstream ML applications.
- Built an AI-powered HS code classification system with asynchronous RESTful API design and Redis caching for real-time categorization with 90% accuracy across 420+ categories, ensuring seamless integration into production systems.
- Developed and deployed a text classification model (Linear SVC) within a FastAPI web application, integrating RESTful APIs to enable real-time, scalable classification of unstructured text data.

Machine Learning Engineer

- Developed reports & dashboards for 6+ international clients in pharmaceuticals, advertising, and academia, leveraging PowerBI and Dash to visualize sales, growth, revenue, and stock KPIs. Integrated advanced data preprocessing, statistical modeling, and classical machine learning techniques to provide actionable insights in weekly/monthly reports.
- Built scalable ETL pipelines connecting Business Central, Xero, and SQL databases with automated daily refreshes, improving data consistency and enabling seamless financial reporting and inventory management for clients.
- Implemented traditional NLP solutions, including TF-IDF, Word2Vec, and Named Entity Recognition (NER), to analyze customer feedback and automate document classification in regulatory compliance tasks.
- Developed and deployed computer vision models using YOLOv4 for real-time object detection, assisting clients in inventory tracking and defect detection in manufacturing/logistics workflows.
- Designed classical ML models (Random Forest, XGBoost, and SVM) for predictive analytics, including sales forecasting and customer segmentation, enabling data-driven decision-making for businesses.

Data Scientist | Business Intelligence Specialist

I am an Artificial Intelligence (AI) Engineer, specializing in designing and implementing modular, scalable, and production-ready AI/ML systems. As a quick learner, I have explored a broad spectrum of areas, including machine learning algorithms, deep learning, natural language processing (NLP), named entity recognition (NER), large language models (LLMs), retrieval augmented generation (RAG), LLM fine-tuning, agentic workflows, computer vision, and data science.

I am dedicated to creating tools for automation and find satisfaction in overcoming challenges, whether collaborating within a team, across teams, or taking ownership of projects.

taplio

The Battle

Adam Sroka

Usman Ali

Followers

Posts

Engagements

Likes

Ready to grow your LinkedIn presence?