Personalized AI

AI that learns each user,
not just the query.

Personalized AI builds a calibration profile for every user across every interaction. Tone, depth, format, and vocabulary adapt automatically to each individual without any configuration on your part.

  • Calibration starts on the first interaction and improves with every session
  • Five independently tracked dimensions per user: tone, depth, format, expertise, vocabulary
  • Works with any model provider. Docklee injects context, the model responds.
ArjunSenior Engineer
Code-firstDense
PriyaProduct Manager
StrategicHigh-level
NoaUX Designer
AnalogiesVisual

Arjuncalibrated

Technical

Sets DOCKLEE_API_KEY, calls .personalise({ userId, query }). Returns PersonalisedContext (see typed interface in SDK reference).

Strategic

Docklee calibrates automatically. Teams ship personalised AI in an afternoon. No training data, no manual fine-tuning needed.

Exploratory

Think of it as a translator between the AI and each person's mental model. It adapts tone, depth, and examples to fit naturally.

Most AI treats every user the same because it doesn't know the difference. Personalized AI builds a calibrated model of each person and makes that the default.

Calibration start

Session 1

Learning begins from the first interaction

Tracked dimensions

5 per user

Tone, depth, format, expertise, vocabulary

Calibration model

Per-user

Isolated profile per user ID, never shared

The problem

Generic AI treats a senior engineer and a new hire identically.

Generic AI

Expert users get over-explained responses they have to skim

Personalized AI

Experts get dense, code-first answers without the primer

Generic AI

Novices receive jargon-heavy replies that lose them immediately

Personalized AI

Novices get plain-language responses with worked analogies

Generic AI

Preferences have to be re-stated in every single session

Personalized AI

Format, tone, and depth are remembered and applied automatically

How it works

Observe. Calibrate.
Personalise automatically.

Step 01

Observe each interaction

After every user interaction, Personalized AI scores the response against five dimensions: expertise level, response depth, tone formality, format preference, and domain vocabulary. Each observation updates the user's profile with a small, confident step.

Step 02

Build and refine the profile

Dimension values accumulate across sessions using a decay-weighted model. Recent interactions count more than old ones. Contradictions are resolved automatically. By session 10, the profile is stable enough to personalise every response without intervention.

Step 03

Inject profile into every response

At query time, the calibrated profile is injected alongside the context bundle. The model receives explicit instructions on tone, depth, format, and vocabulary for this specific user. No prompt engineering required on your end.

Profile calibration, real-time
Active

Arjun, Senior Engineer

First interaction

learning

Expertise level18%

Response depth22%

Tone formality15%

Format preference20%

Domain vocabulary12%

Capabilities

Personalisation that actually works in production.

Continuous calibration

The profile improves with every session. Dimension values converge toward the user's true preferences over time using a decay-weighted model that prioritises recent interactions.

Five tracked dimensions

Expertise level, response depth, tone formality, format preference, and domain vocabulary are tracked independently. Each dimension calibrates at its own pace.

Per-user isolation

Every profile is scoped to a single user ID and stored in isolation. Profiles never influence each other. Multi-tenant architectures are supported natively.

Session-to-session memory

Calibrated dimensions persist across sessions. A user who preferred concise answers last week gets concise answers this week without re-explaining anything.

Override controls

Users or operators can pin specific dimensions to fixed values. Pinned dimensions are never adjusted by the calibration system, giving humans full override capability.

Profile inspection API

Read any user's current profile at any time. Use dimension values to drive product decisions, A/B tests, or custom UI experiences beyond just AI responses.

API

One call. Calibrated response.
No configuration needed.

Pass a userId and a query. Personalised AI handles calibration, profile injection, and model routing automatically.

  • JavaScript and Python SDKs
  • REST API with profile inspection
  • Streaming completions supported
  • Works with GPT-4o, Claude, Gemini
Live
import { Lynk } from "docklee";
const lynk = new Lynk({ memorySpaceId: "space_app" });
// Recall this user's calibrated preferences
const prefs = await lynk.recall("expertise, tone, and depth", {
endUser: "user_7x91k",
topK: 5,
});
// Shape the response for this user with any model
const result = await openai.chat.completions.create({
model: "gpt-4o",
messages: [
{ role: "system", content: personalize(prefs) },
{ role: "user", content: "How do I configure the SDK for production?" },
],
});
// Adapts automatically to the user's expertise, tone, and depth

Use cases

Every AI product improves when it knows its users.

Personalized AI is the layer that turns a one-size-fits-all AI into something that feels like it was built for each person.

AI copilots

In-product copilots that respond to a senior engineer in concise code blocks and to a junior team member in plain-language walkthroughs, automatically.

Support chatbots

Support bots that calibrate over time to each customer's language, technical level, and preferred response format, reducing handling time and escalation rate.

Content and recommendation

Recommendation engines that adapt their explanation style per user. An expert gets a data-dense brief. A novice gets a narrative. Same AI, calibrated output.

Enterprise search

Search interfaces that learn each person's preferred result depth, summary style, and source preference. The search experience improves just by being used.

Personalized AI

Build AI that gets better
the more it knows each person.

Calibration starts from the first session. No training data, no prompt engineering. Just AI that improves with every interaction.