Machine Learning Foundations
A decision-centered introduction to supervised learning: frame prediction problems, trace how models learn, compare model biases, evaluate generalization honestly, and diagnose what to improve next.
Ai Ml And Agentic Systems
TRACKS
Deep Learning and Neural Networks
How neural networks build representations, receive credit through gradients, train reliably, encode architectural bias, transfer knowledge, and cross the boundary into inference systems.
LLM Foundations
Transformer-era language modeling concepts, architectures, and capability framing.
LLM Training, Alignment, and Serving
Training data, distributed pretraining, post-training, alignment loops, inference optimization, and production serving for large language models.
RAG, Agents, and LLM Production
Evidence-grounded LLM product architecture: RAG pipelines, agent-facing context, evaluation slices, observability, cost, safety boundaries, and production release judgment.
Multimodal Foundations and Vision-Language Models
Multimodal encoders, contrastive learning, grounding, and vision-language model design.
Reasoning and Frontier Post-Training
Synthetic data, verifiers, process supervision, reasoning traces, and frontier post-training loops.
Agent Runtime and Tool-Use Systems
Tool schemas, planners, sandboxes, browser agents, orchestration loops, and runtime failure handling.
Agent Memory, Planning, and Context Engineering
Design memory, context assembly, and planning state for long-lived LLM agents that must preserve evidence, recover work, and act under operational constraints.
World Models and Model-Based Agents
Latent dynamics, predictive state, imagination-based planning, and model-based agent architectures.
Diffusion, Audio, Video, and Generative Media
Diffusion models, latent media generation, controllability, and evaluation across image, audio, and video systems.
Agent Safety, Guardrails, and Policy Systems
Guardrails, policy enforcement, action filtering, runtime controls, and the trust boundaries needed around agentic systems.
LLM Evaluation and Benchmarking
Offline evals, task suites, judge systems, reliability trade-offs, and the measurement discipline required to compare LLM behavior honestly.
Multi-Agent Systems and Coordination
Coordination protocols, role assignment, negotiation, and the design patterns for systems composed of multiple autonomous agents.
Tool Learning and Environment Interaction
Tool selection, environment feedback, learned interaction policies, and the mechanisms that let agents improve through action.
Causal ML and Decision Making
Interventions, counterfactual thinking, uplift, and the use of causal structure to support better decisions than prediction alone.
Optimization, Generalization, and Learning Theory
Bias-variance trade-offs, sample complexity, optimization behavior, and the theory that explains why learning succeeds or fails.
ML Systems and Training Infrastructure
Draft track for feature platforms, training pipelines, experiment systems, model deployment, and inference operations.
Probabilistic Modeling and Bayesian Inference
Latent-variable models, priors, posterior reasoning, and the probabilistic view of learning under uncertainty.
Reinforcement Learning and Sequential Decision Making
Value functions, policy learning, exploration, planning, and the algorithms for acting under delayed feedback.
Representation Learning and Self-Supervision
Embeddings, contrastive objectives, pretext tasks, and the training recipes that build reusable latent structure from raw data.
Robotics, Automation, and Embodied AI
Sensors, actuators, frames, control loops, perception, planning, safety, and simulation-to-reality gaps in embodied systems.
MLOps, DataOps, and Model Operations
Operational discipline for machine learning: datasets, training pipelines, evaluation, deployment, monitoring, drift, lineage, and rollback.
AI Research Literacy and Paper Reading
Read AI research with discipline: claims, baselines, ablations, datasets, benchmarks, limitations, replication, and implementation judgment.
Human-AI Collaboration and Agentic Workflows
Design reliable work with AI assistants and agents: delegation, context, review, tool boundaries, memory, failure recovery, and human judgment.