Soft Tech Group
Research & innovation

Advancing AI & predictive analytics for tomorrow.

We continuously invest in AI research, enterprise architecture, and intelligent systems, enabling our clients to benefit from proven innovations and capabilities validated through practice, experience, and real-world application.

0
Published papers
0
Research domains
0
Models benchmarked
0
Applied AI projects
Latest publications

Peer-reviewed research & applied findings.

Research - Informed Intelligence and practical Insights developed through Enterprise Engineering, Innovation, and Real-Industry Experience.

PredictiveJul 2026 · 12 min read

Evaluating Foundational Predictive Models: A Rigorous Experimental Framework for Qualitative and Quantitative Data Modalities

AbstractA Unified Experimental Framework for Comparative Analysis of Foundational Predictive Models Across Classification and Regression Predictive modeling serves as a cornerstone of data-driven decision-making across industries, yet comparative analyses often remain confined to singular task types, limiting insights into crossdomain model generalizability. This research addresses this critical gap by introducing a comprehensive, unified framework for the rigorous evaluation of five foundational model categories. The study concludes that model performance is inherently context-dependent.

Predictive ModellingModel EvaluationMachine LearningDeep LearningSpecialized Models
Read paper
LLMMay 2026 · 18 min read

Advancing Large Language Model Reasoning Techniques: Methods Enabling LLMs to 'Think' Beyond Text Generation for Reliable and Explainable AI

AbstractLarge Language Models (LLMs) have revolutionized artificial intelligence applications, extending from writing assistants to Retrieval-Augmented Generation (RAG) systems. However, understanding how LLMs "reason" process complex queries and generate reliable results beyond mere text generation, it has become a pivotal research focus. This paper surveys the core reasoning techniques that empower LLMs to simulate logical thinking: Chain-of-Thought (CoT), Self-Consistency, ReAct (Reason + Act), and Plan-and-Solve Reasoning.

ReasoningChain-of-ThoughtHallucination MitigationExplainabilityTrustworthy AIReAct
Interactive benchmark
LLMNov 2025 · 14 min read

Maximizing Scalable AI: Efficient Language Model Adaptation Using Fine-Tuning, Direct Preference Optimization, and Online Reinforcement

AbstractOptimizing both large language models (LLMs) and small language models (SLMs) for realworld use requires thoughtful post-training adaptation. This overview highlights three key strategies: Supervised Fine-Tuning, Direct Preference Optimization (DPO), and Online Reinforcement Learning. Supervised Fine-Tuning refines pre-trained models using labeled, instruction-following datasets. This improves task accuracy and response controllability by aligning outputs with ground. By aligning model behavior with user intent while addressing bias and training inefficiencies, they significantly improve language model utility in real-world applications.

Fine-tuningDPORLHFReinforcement LearningModel Adaptation
White paper
Business AISep 2025 · 9 min read

Improving the Accuracy and Interpretability of Sales Forecasts in ERP Systems using Predictive Analytics with Machine Learning and Large Language Models

AbstractThe industry's efficiency hinges on the seamless integration of production, inventory, and distribution processes. Enterprise Resource Planning systems are the cornerstone of this integration, managing core functionalities from inventory to customer data. The next evolutionary step for these systems is the integration of Predictive Analytics, powered by Machine Learning (ML) and, more recently, Large Language Models (LLMs). This paper investigates the confluence of these technologies to enhance sales forecasting within ERP systems.

ERP SystemsSales ForecastingPredictive AnalyticsLLM IntegrationInterpretability
Case study
How we evaluate enterprise AI

From research question to production.

Every insight we publish is earned through rigorous research, enterprise engineering, validation processes that guide the architectural design, implementation and deployment of client solutions.

1

Research question

Frame the business problem, establish measurable outcomes, identify the highest-value AI opportunity.

2

Data engineering

Acquire, integrate, govern, and prepare enterprise data for scalable AI development.

3

Benchmark models

Evaluate, and optimize AI models against enterprise performance and accuracy.

4

Evaluation

Access reliability, security, explanability, compliance, latency, and operational impact before deployment.

5

Production validation

Stress-test AI systems under real-world workloads, governance controls, monitoring and edge-case scenarios.

6

Deployment

Continously monitor and capture feedback, and continously optimize models for sustained business value.

Current research focus · 2026

Where our attention is right now.

Agentic AI
Robotics
Physical AI
Predictive Intelligence
Computer Vision
AI Governance
Edge AI
Multimodal Systems
Research areas

Where we focus.

Your next advantage

Your next competitive advantage is already in your data.

Let's identify where AI creates the greatest measurable value across and engineer it into production.