Sharing practical tips and the raw truth while navigating the intersection of AI , business , and UX . do follow for more insights.
agentic pipeline
Joined October 2025
- Tweets124
- Following4
- Followers91
- Likes32
From today i am going to share the knowledge on how to build an AI system :
Conceptual prerequisites the readers should understand by the end of the series:
1) What an AI agent is, and how it’s different from a simple chatbot.
AI product manager retweeted
chest code to understand agentic AI codebase ?
- learn parsing and processing difference
- learn the workflow between parsing and processing
- learn about the python data structures (bcz it is very helpful to understand shared state between agents)
- learn about the different messages works in an agent (bcz you have to know which data to be shared between the agent to prevent context rot)
chest code to understand agentic AI codebase ?
- learn parsing and processing difference
- learn the workflow between parsing and processing
- learn about the python data structures (bcz it is very helpful to understand shared state between agents)
AI product manager retweeted
2/n
- python function used as tool with speicific name and description
- the tool may do just parsing of skill md file
- skill md should contain name and description to easily scan to avoid latency
AI product manager retweeted
rules of thumb for building agentic AI 1/n
- scratchpad is specific to that agent , mostly contains ReAct messages.
- scratchpad final information is transferred to the state for state management
- that final info is what is needed for the future agents
2/n
- python function used as tool with speicific name and description
- the tool may do just parsing of skill md file
- skill md should contain name and description to easily scan to avoid latency
rules of thumb for building agentic AI 1/n
- scratchpad is specific to that agent , mostly contains ReAct messages.
- scratchpad final information is transferred to the state for state management
- that final info is what is needed for the future agents
AI product manager retweeted
today i will post about- Building Agent Tools: APIs & Resiliency
1) Concept: Why Tools Matter
Foundation models are pattern‑predictors; they cannot see live data or take real actions without tools.
Tools are the bridge from static training to live systems: they let the agent
AI product manager retweeted
if you take this series as a base to your agentic ai curiosity, no one stops you in that journey
lets talk about database memory and chat interface today:
1) Concept: Sessions vs Long‑Term Memory
A session is a chronological log of one continuous conversation: events + state.
AI product manager retweeted
hi guys, hope your are enjoying the agentic ai series.
today in day4, let me share about importance of context engineering.
1) Concept: Context Engineering
Context engineering is how you dynamically build the model’s input each turn: system instructions, tools, history, RAG
AI product manager retweeted
in day3 lets talk about wiring Your Agent Pipeline using langgraph framework
1) Concept: Graph‑based Orchestration
You use a StateGraph (LangGraph) to describe nodes (agents/steps) and edges (transitions) forming a DAG (Directed Acyclic Graph).
AI product manager retweeted
After creating the first AI agent in ReAct pattern in day1,
Let's talk about the Multi‑Agent Architecture & Role Separation in Day2
1) Concept: Core Agent Architecture
Every agent you build follows three components:
Model (Brain): Does the reasoning and planning.
AI product manager retweeted
let me start with DAY 1:
1) Concept : What is an AI Agent?
An AI agent is a system built around a language model that can plan and take actions with tools to achieve a goal without constant human direction.
decisions are paused for human approval or correction
**Evaluation‑Gated Deployment** – Only deploying agent versions that pass predefined evaluation suites and safety checks.
**Vertex AI Agent Engine** – Managed Google Cloud service for hosting, scaling, and observing agents.
**LLM‑as‑a‑Judge** – Using a separate language model to grade agent outputs/trajectories against a rubric.
**Golden Set** – Small, human‑curated set of labeled examples defining “good” behavior used for evaluation.
**Human‑in‑the‑Loop (HITL)** – Pattern where critical
series of logs forming the full trajectory of one task
**Metric** – Aggregated numeric measure of behavior or performance (latency, error rate, task success).
**Agent Quality** – Multi‑dimensional measure including Effectiveness, Efficiency, Robustness, and Safety \& Alignment.