jnaapti

Releasing soon A hands-on jnaapti workshop

Agentic Engineering: Rethinking Knowledge Work

From endless prompts
to effective automation.

Learn to automate everyday work with AI—without giving up control.

Express interest Explore the curriculum
  • Ground answers in your data
  • Turn repeatable work into skills
  • Build in human oversight

For software engineers & technical knowledge workers.

FROM INTENT TO OUTCOME WORKFLOW / 01
THE STARTING POINT

“Help me get this work done.”

GIVE IT THE RIGHT CONTEXT
Your dataYour toolsYour rules
01Plan
02Act
03Verify
Human review where it matters
A repeatable workflow.Not another one-off answer.
Rules for the predictable.AI for the ambiguous.
20 hoursFocused, practical learning
12 modulesFoundations to automation
7 phasesOne connected learning path
Hands-onLearn by building

A different way to work

A good answer is a start.
A working system is the goal.

AI can write code, summarize documents and suggest a plan. But disconnected prompts still leave you moving context, checking results and repeating the same steps.

This workshop connects the dots: the models, the data, the tools, and the engineering practices that turn AI into a reliable part of your workflow.

Less “prompt, wait, repeat.”
More “tell me when you need my attention.”

Learn it. Build it. Put it to work.

Three shifts. Real engineering depth.

Go beyond tool demos. Learn the principles, then apply them in hands-on labs.

Give AI the context
it actually needs.

Build retrieval-augmented generation (RAG) pipelines with embeddings, chunking, vector indexes and knowledge graphs. Explore a fully local pipeline with Chroma.

RAGKnowledge graphsLocal retrieval

Turn your process
into reusable skills.

Combine scripts with agents. Connect tools through MCP, automate a browser with Playwright, and map human checkpoints across the software development lifecycle.

Agent skillsMCPSDLC

Automate the work.
Keep the judgment.

Design stateful workflows with security, permissions and cost governance. Progress to event-driven agents with policy gates, provenance and recoverable execution.

GuardrailsHuman reviewEvent-driven agents

The learning journey

From understanding AI
to engineering with it.

Seven connected phases. Open a phase to see the modules and topics.

01Understand the shiftKnow what AI can do—and where engineering judgment still matters.2 hours · 2 modules

Foundations of AI and Machine Learning

1 hour

AI, ML and neural networks; symbolic vs neural AI; tacit knowledge; training, inference and memory limitations.

Generative AI and the Deterministic-Stochastic Shift

1 hour

Rules vs probabilistic output; generative architectures; context windows; knowledge work as input, process and output; structured cross-domain synthesis.

02Look inside the engineUnderstand how a model produces answers and how an agent takes action.2 hours · 1 module

LLM Internals: Parameters, Sampling, and Tool Calling

2 hours

Tokens, logits and probabilities; temperature, Top-K and Top-P; reasoning effort; JSON-schema tools; the agent loop; MCP discovery; structured output.

03Connect AI to your knowledgeBuild retrieval pipelines that bring private knowledge into the conversation.3 hours · 2 modules

Retrieval, RAG, and Knowledge Graphs

1.5 hours

Keyword, semantic and structured retrieval; ingestion-to-generation pipelines; triples, ontologies and RDF/OWL; Neo4j; GraphRAG and Agentic RAG.

RAG Deep Dive: Embeddings, Chunking, and Vector Indexes

1.5 hours

Embeddings and cosine similarity; FAISS vs Chroma; chunking; iterative retrieval; topic clustering; cross-encoder reranking; a fully local RAG pipeline.

04Build agents that do real workTurn repeatable work into reusable skills, with humans at the right checkpoints.4 hours · 2 modules

Agent Architecture: Harnesses, Providers, and Models

1.5 hours

Model, provider and harness layers; sessions and sandboxes; skill discovery; permissions; human oversight; choosing agents vs scripts.

Building Agent Skills: Design and Workflows

2.5 hours

SKILL.md; parameterized scripts; hybrid workflows; plan verification; code mode; dossier patterns; critical-review sub-agents; Playwright MCP; SDLC human touch points; Sprint velocity, DORA, cycle time and defects.

05Make automation accountableManage context, state, security and cost before you hand over control.3 hours · 2 modules

Production Agent Workflows: Context, State, and Automation

1.5 hours

PARA organization; data sources; knowledge-work event logs; state transitions; signal collectors; task granularity; engineering, AI and institutional knowledge.

AI Security, Governance, and Adoption Guardrails

1.5 hours

Prompt injection and OWASP LLM risks; least privilege, zero trust and ZDR; containment; access control; compliance; cost governance; approved-model policies; responsible AI.

06Engineer beyond the chat windowBuild stateful, multi-provider workflows and explore local model deployment.4 hours · 2 modules

LangChain, LangGraph, Deep Agents, and DSPy

2.5 hours

Tools and typed messages; Pydantic output; conversational memory; state graphs, loops and interrupts; streaming and subgraphs; planning and context compression; DSPy signatures, modules and optimizers.

Local Models and the Future

1.5 hours

Hosted, low-cost API and self-hosted deployment; GPU economics; GGUF quantization; Ollama and HuggingFace.

07Move from reactive to proactiveDesign workflows that respond to events—not just to your next prompt.2 hours · 1 module

Proactive Agents and Event-Driven Workflows

2 hours

Autonomy vs agency; objects, events and behaviours; ActiveGraph; policy gating; provenance and replay; hybrid DAGs; Prompt-to-DAG; checkpoints, recovery and parallel agents.

Inside the toolkit

  • OpenCode
  • Playwright MCP
  • Chroma
  • Neo4j
  • LangChain
  • LangGraph
  • Deep Agents
  • DSPy
  • Ollama
  • HuggingFace

Is this for you?

For people who want
to build, not just browse.

Designed for software engineers and technical knowledge workers who want to make AI part of a real process.

  • You write software and want agents to help across the SDLC.
  • You work with documents, data and tools—and want to connect them.
  • You need to understand the trade-offs before automating decisions.

What to bring

  • Knowledge of a programming language, preferably Python.
  • Basic command-line and version-control experience.
  • Familiarity with coding agents such as Claude Code is helpful.

Your lab setup

A modern laptop, browser and VS Code. Plan for 16 GB RAM and 20 GB disk space, plus OpenAI API and OpenCode Go access.

Local-model experiments use Ubuntu 24.04 or macOS and a GPU. Lab VMs for open-weight experimentation are available at additional cost.

Your instructor

Gautham Pai

Founder, jnaapti

21years in engineering
15years teaching

Taught by someone who
builds the systems.

Gautham approaches knowledge work as an engineering problem: where does knowledge live, how does it move, and which steps need human judgment?

His journey spans IBM, browser-based IDEs, and production semantic-web platforms. At jnaapti, he built Virtual Coach, Cloudlab and IntelliRead—the learning, lab and knowledge systems behind his own agent workflows.

The progression in this workshop is not theoretical. His training workflows—from preparation and content generation to delivery and follow-up—run on the same systems he builds for clients.

75+ organizations. 100 technologies.

jnaapti’s technical training experience includes Adobe, PayPal, Intel, Visa and Morgan Stanley.

Before you dive in

Good questions.
Clear answers.

When does the workshop start?

Releasing soon. Dates are not announced yet. Use the interest enquiry to ask about launch details.

Do I need to be an AI expert?

No. The program starts with AI and machine-learning foundations. You should know a programming language, preferably Python, and be comfortable with the command line and version control.

Is this only about AI coding assistants?

No. Coding agents and SDLC workflows are part of the program. You also learn retrieval, knowledge graphs, reusable skills, security and event-driven automation for broader technical knowledge work.

Will I build things, or just watch demos?

The 20-hour program includes hands-on labs. The outline covers local RAG, agent skills, browser automation, programmatic agent frameworks and local-model experiments.

What about access, pricing and lab costs?

Workshop pricing and enrollment details are not announced yet. The outline requires OpenAI API and OpenCode Go access. Optional lab VMs for open-weight experimentation cost extra. Ask us about setup and access before enrolling.

Releasing soon

Your next step isn’t another prompt.
It’s a better way to work.

Explore agentic engineering with jnaapti.
Tell us you’re interested, and ask about the upcoming workshop.

Express interest in the workshop

An enquiry, not an enrollment. No payment required.