Aurimas Griciūnas's LinkedIn Strategy…

TOON (Token-Oriented Object Notation) is out for some days now and it aims to make communication with LLMs more accurate and token-efficient. The TOON topic is now one of the hottest news on the LLM…

LinkedIn post image: TOON (Token-Oriented Object Notation) is out for some days now and it aims to make communication with LLMs more accurate

58876829 viral

AI10 months ago

View on LinkedIn

You must know these 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗦𝘆𝘀𝘁𝗲𝗺 𝗪𝗼𝗿𝗸𝗳𝗹𝗼𝘄 𝗣𝗮𝘁𝘁𝗲𝗿𝗻𝘀 as an 𝗔𝗜 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿. If you are building Agentic Systems in an Enterprise setting you will soon discover that…

LinkedIn post image: You must know these 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗦𝘆𝘀𝘁𝗲𝗺 𝗪𝗼𝗿𝗸𝗳𝗹𝗼𝘄 𝗣𝗮𝘁𝘁𝗲𝗿𝗻𝘀 as an 𝗔𝗜 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿.

38737726 viral

AI9 months ago

View on LinkedIn

𝗔𝗜 𝗢𝗯𝘀𝗲𝗿𝘃𝗮𝗯𝗶𝗹𝗶𝘁𝘆 is a must have in your tool belt as an AI Engineer. 𝗧𝗿𝗮𝗰𝗶𝗻𝗴 sits at the core of it, why is it important? Tracing and instrumentation of software have been aroun…

LinkedIn post image: 𝗔𝗜 𝗢𝗯𝘀𝗲𝗿𝘃𝗮𝗯𝗶𝗹𝗶𝘁𝘆 is a must have in your tool belt as an AI Engineer. 𝗧𝗿𝗮𝗰𝗶𝗻𝗴 sits at the core of i

36066506 viral

AI Engineering9 months ago

View on LinkedIn

I have been developing Agentic Systems for the past few years and the same patterns keep emerging. 👇 𝗘𝘃𝗮𝗹𝘂𝗮𝘁𝗶𝗼𝗻 𝗗𝗿𝗶𝘃𝗲𝗻 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 is the most reliable way to be successfu…

LinkedIn post image: I have been developing Agentic Systems for the past few years and the same patterns keep emerging. 👇

36135676 viral

AI9 months ago

View on LinkedIn

𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗥𝗔𝗚 and what you need to know about it as an AI Engineer? Simple naive RAG systems are rarely used in real world applications. We are usually adding some agency to the RAG system -…

LinkedIn post image: 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗥𝗔𝗚 and what you need to know about it as an AI Engineer?

35142405 viral

AI9 months ago

View on LinkedIn

Building even a simple 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻 𝗴𝗿𝗮𝗱𝗲 𝗥𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 𝗔𝘂𝗴𝗺𝗲𝗻𝘁𝗲𝗱 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻 (𝗥𝗔𝗚) 𝗯𝗮𝘀𝗲𝗱 𝗔𝗜 𝘀𝘆𝘀𝘁𝗲𝗺 is a challenging task. Read until the end to und…

LinkedIn post image: Building even a simple 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻 𝗴𝗿𝗮𝗱𝗲 𝗥𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 𝗔𝘂𝗴𝗺𝗲𝗻𝘁𝗲𝗱 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻 (𝗥𝗔�

32028515 viral

AI9 months ago

View on LinkedIn

Topics & Content Focus

Primary Topics

Practical LLM system architecture and context engineering (RAG, CAG, agent memory, chunking, retrieval design)Agentic systems and autonomous AI workflows (Agentic RAG, query analysis, tool routing, planning loops)Productionization, reliability and governance of GenAI systems (observability, tracing, evaluation, RBAC, guardrails)

Secondary Themes

Vector search and embedding strategy (model choice, similarity measures, hybrid search, metadata design)Prompt and system design for complex AI applications (system prompts, tool registries, procedural memory)Cost and performance trade-offs for enterprise GenAI (token economics, latency bottlenecks, model selection)

Industry Focus

Enterprise GenAI platforms and AI infrastructureB2B AI product engineering and applied ML/LLM teamsTechnical AI leadership for production-grade LLM applications

Content Categories

Technical architecture breakdowns and system design blueprintsBest-practice frameworks and taxonomies (types of memory, types of retrieval, span/trace definitions)Implementation checklists and tuning levers for AI engineersOpinionated guidance on simplifying over-engineered agentic/RAG systems

Performance Insights

403.2%

Avg Engagement Rate

STABLE

Performance Trend

Best Performing Topics

AI observability and tracing for RAG/agentic systemsAgentic RAG pipelines with minimal necessary agency and reranking focusHidden complexity and design choices in production-grade RAG architectures

Virality Signals

High share counts on posts that map the full system and enumerate concrete tuning levers (observability, RAG architecture)Above-average engagement when content deconstructs real production complexity rather than focusing on narrow tricksConsistent comment activity driven by explicit invitations for practitioners to share experiences and ‘war stories’Technical depth presented in a highly structured, skimmable format that lends itself to saving/sharing as a reference

Structure & Quality

400

Avg Length (Words)

HIGH

Depth Level

ADVANCED

Expertise Level

0.85/10

Uniqueness Score

Common Hooks

Concept reframe hooks that define or rename a key component (e.g. positioning agent memory as the core of context engineering)Difficulty/complexity hooks that reveal hidden challenges (e.g. why production-grade RAG is hard)‘What you need to know as an AI Engineer’ hooks that explicitly target a practitioner personaTrend-plus-depth hooks combining hyped ideas (CAG, Agentic RAG) with sober implementation detail

Common Endings

Open question inviting opinions or experiences (e.g. ‘What are your thoughts on X?’)Request for war stories or practical lessons from the fieldDirect invitation to comment with experiences or resultsOccasional teaser of future, deeper posts (e.g. ‘More on this in future posts, stay tuned!’)

Value Delivery Methods

Transforms vague buzzwords (Agentic RAG, memory, CAG) into concrete models, taxonomies and step-by-step flowsProvides implementation-level detail for AI engineers, including architectural choices, trade-offs and caveatsSurfaces under-discussed production concerns (observability, RBAC in caches, cost tracing, evaluation at span level)Offers opinionated simplification guidance, helping readers avoid overbuilding agentic pipelines

Formatting Style

Bolded or stylized headline phrase at the top (often using Unicode bold)Sectioned structure with mini-headings (Definitions, Retrieval, Generation, Considerations, Why it matters)Numbered lists that read like procedural checklists or pipelinesBulleted sub-points with arrows and checkmarks/crosses to signal do/don’t guidanceOccasional use of emojis and symbols to visually segment pros, cons and notes

Audience & Tone

YES

Question Usage

0.5%

Response Rate

Detected Tone

Practitioner-scholar (mix of hands-on detail with conceptual framing)Systems thinker focused on pipelines, topology and moving partsTechnical yet approachable, assuming competence but explaining clearlyOperationally opinionated, especially about simplification and production realitiessemi-formalsecond-person

Interaction Style

Peer-to-peer knowledge exchange with other practitioners and technical leadersCollaborative framing that invites others to validate, challenge or extend the proposed frameworksScenario-based prompts encouraging readers to map advice to their own production systemsCommunity-building around ‘AI engineer as system designer’ identity rather than around tools alone

Community Building Signals

Regular use of inclusive language (‘we’, ‘as AI Engineers’) to create a shared professional identityEncouraging comment threads that function as a repository of additional tactics and war storiesPositioning posts as living artefacts (frameworks and checklists) that practitioners can iterate on together

Writing Style Patterns

Content Strategy

Hook: Concept reframe hooks that define or rename a key component Tone: semi-formalCTA: Open-ended discussion prompts aimed at peers (‘Let

Writing style breakdown

- Core style

Professional, educational, and highly structured.

Conversational but not casual; "LinkedIn thought-leadership for engineers" vibe.

Very didactic: posts feel like mini-guides, not opinions or hot takes.

Persuasive mainly through clarity and structure, not hype or emotion.

- Formal vs casual

Language: semi-formal. Uses domain-accurate technical terminology (RAG, CAG, vector DBs, MLOps, observability, MCP, etc.) with almost no slang.

Contractions are common ("Let’s", "don’t", "you’ll") which softens the tone.

No profanity. No sarcasm. No edgy humor.

- Emotional tone and energy

Medium energy: confident and engaged, but calm and measured.

Tone is optimistic and constructive ("you should strive", "it is really easy to start prototyping").

No dramatization or outrage; excitement is expressed via "this is important", not via hyperbole.

- Starts with a strong, compact hook, often

A concept + why it matters ("AI Agent’s Memory is the most important piece...")

A question targeted at a role ("what you need to know about it as an AI Engineer?")

A bold assertion + "read until the end" incentive.

- Heavy use of

Definitions ("Few definitions:", "A short summary:")

Explicit enumerations ("It is useful to group the memory into four types:")

Checklists and step-by-step processes.

Occasional rhetorical questions in the body ("What might a trace look like...?") but most questions are reserved for the CTA.

- Very frequent meta-signposting phrases

Let’s look into...

Let’s explore...

This is where:

For this you need to understand...

Uses emojis and icons systematically (✅, ❌, ❗️, ℹ️, ➡️, 👇) for structure and emphasis, not for humor.

- Addressing the reader

Primary person: second person ("you", "you must know", "you should strive").

- First person used sparingly and mainly for

Expertise anchoring: "I have been developing Agentic Systems..."

Opinion framing: "My Honest Thoughts:", "Why I think so:"

Almost no third-person storytelling; focus is on "you as AI/MLOps/AI Engineer".

- Both are used, but phrased politely

Direct: "You must know these...", "Always try to solve...", "Be very careful about what you cache..."

Softened suggestions: "You should strive to provide...", "Think about an assistant that...", "Consider this as..."

Often frames advice as "you should" or "it is useful to" rather than pure orders.

Aurimas Griciūnas

Warm Analysis

Performance Overview

Top Posts by Engagement

Posting Patterns & Frequency

Best Performing Days

Best Performing Times To Post