myAInotes
Building Intelligent Systems That Learn, Reason and Evolve
Researcher | AI Engineering | Computational Finance | Algorithmic Trading
Welcome to my AI Notes.
I'm Shyaam Prasadh, an AI researcher and engineering leader working at the intersection of artificial intelligence, reinforcement learning, agentic systems, statistical modelling, computational finance, and decision intelligence.
Over the past decade, I've worked across quantitative finance, enterprise AI, machine learning research, and large-scale decision systems. My work spans building autonomous AI agents, developing statistical models for financial markets, algorithmic trading, and sports betting, designing enterprise AI platforms, and applying machine learning to complex real-world decision problems.
Today, I lead Enterprise AI Engineering at Entain, where I build large-scale AI systems, LLM-powered products, and intelligent agent architectures deployed across multiple business domains.
This website serves as my personal research notebook, documenting ideas, experiments, papers, and engineering lessons from building intelligent systems that continuously improve over time.
"How do we build intelligent systems that continuously learn, reason, adapt, and evolve?"
Research Interests
- Agentic AI and autonomous systems
- Reinforcement learning and sequential decision making
- Multi-agent systems and collective intelligence
- Large language models and post-training
- Statistical modelling and probabilistic inference
- Computational finance, quantitative research, and algorithmic trading
- Sports betting analytics and decision intelligence
- Agent memory, reasoning, and planning
- AI observability, evaluation, and governance
- Enterprise AI platforms and production systems
Current Missions
What I'm building right now
Prometheus
Enterprise AI PlatformTrading, risk, and operational intelligence powering Entain's global decision-making infrastructure.
EntainGPT
25,000 Employees · 250M+ TokensSuite of LLM-powered services for compliance, operations, and enterprise knowledge with multi-million GBP impact.
AgenticOps
Autonomous Incident ManagementSelf-healing systems with hierarchical memory, safety filters, and sub-3s latency for production reliability.
Self-Evolving Agents
Trading · Data Science · Decision IntelligenceAgents that learn continuously, adapt reasoning, and operate as scalable autonomous organizations.
Current Research
Although my interests appear diverse, they are connected by a common objective: building systems that make better decisions under uncertainty.
Whether modelling financial markets, optimizing sports betting strategies, designing enterprise AI platforms, or developing autonomous agents, the underlying challenge remains remarkably similar. Intelligent systems must observe their environment, learn from experience, reason about uncertainty, make decisions, and continually improve through feedback.
The next generation of AI will not simply consist of larger language models or better copilots. It will comprise self-evolving agent ecosystems: interconnected networks of specialised agents that continuously observe, reason, act, evaluate, and improve.
In an enterprise setting, this means:
- Trading agents that adapt to changing market conditions
- Risk agents that learn from fraud patterns and operational incidents
- Personalisation agents that optimise customer experiences
- Knowledge agents that maintain organisational memory
- Planning agents that coordinate enterprise workflows
- Governance agents that monitor safety, compliance, and performance
Building these systems requires advances across reasoning, memory, reinforcement learning, planning, observability, governance, statistical modelling, and systems engineering. It is not a model problem alone: it is a systems engineering challenge.
Recent Blog Posts
- Implementing Adaptive Compute in Prometheus: Learning When LLMs Should Think Reasoning HTML
- General Agents as World Models Agents PDF
- Small LLMs as Agents Agents PDF
- Misinterpretability of Illusion in the Reasoning of LRMs LLMs PDF
- Model Agnostic Meta Learning Deep Learning
- Reasoning via Internal Rewards LLMs
- Faithful CoT LLMs
- Much Ado About Proofing: DeepSeek Prover-V2 LLMs
- Why Do Multi-Agent LLM Systems Fail Agents
- Agentic Self-Awareness Agents
- Investigation of R1 Zero-like Training RL Training
- Future of Agents Agents
- LLM Powered Autonomous Agents Agents
- Scaling Laws Summary Scaling Law
- Quantization of LLMs LLMs
- Curated List of RAG Papers Generative AI
- Pareto Alignment via Preference Adaptation for LLMs AI Alignment
- Evaluator's Reading List AI Alignment
- Intriguing Properties of NNs Deep Learning
- Beyond Preferences in AI Alignment AI Alignment