Active Beta · USMLE & MBBS · 1,247 students

The end of re-explaining yourself to AI.

Stella is the clinical reasoning co-pilot built for medical students — with a 3-tier memory architecture that remembers your patient cases, tracks your weak topics across sessions, and never throttles you mid-differential.

Stella app — full dashboard with persona selector, session list, and personalized greeting

↑ The actual product. Your name, your streak, your last session — right where you left it.

Scroll
1,247 active students 52,418 sessions logged 1.8M MCQs generated 89% retention rate 4.9★ avg rating 23 universities 0 sessions lost to throttling 3-tier memory 4 response modes 1,247 active students 52,418 sessions logged 1.8M MCQs generated 89% retention rate 4.9★ avg rating 23 universities 0 sessions lost to throttling 3-tier memory 4 response modes
01
01 — The problem

You're not studying wrong.
You're studying with tools built for someone else.

Generic chatbots were designed for emails and essays. When you bring them a 2am clinical vignette with three distractors and a side of pathophysiology, they give you a paragraph and a shrug.

01

Context amnesia

You re-explain your patient case from scratch every session. The chatbot has no idea you already covered the RTA Type 1 mechanism three messages ago.

02

Throttled mid-differential

Commercial caps shut you down at the worst moment — right when you're 12 sessions deep and finally cracking a high-yield case.

03

Answers, not thinking

You don't just need the right answer. You need to understand why the A is right, why B is a trap, and how to walk that logic in an oral exam.

02 — Meet Stella

A clinical reasoning co-pilot
not another chatbot.

Stella behaves like a senior resident sitting next to you. Three layers of memory, four response modes, four exam-specific personas, and a Compaction Engine that turns every session into the foundation for the next one.

03
03 — The architecture

Three tiers of memory.
Zero context drift.

Stella's L1 / L2 / L3 architecture scope-partitions your knowledge state — so your next session begins exactly where the last one ended.

L1 · Active

Live conversation

The immediate buffer. Holds the active patient history, the labs you just pasted, the mechanism you're tracing, the distractors you're ruling out. Evolves in real time as you think out loud.

Scope: Current session
L2 · Knowledge state

Your weak map

Stella parses every exchange to classify each topic as Weak, Needs Work, Strong, or Neutral. Routes your next review, your next MCQ, your next flashcard to where it matters most.

Scope: All sessions · Updated live
L3 · Compaction

Permanent digests

At session end, the Compaction Engine distills everything you covered into a chronological ledger. Next time you open Stella, she loads your L3 — and you're back in context within seconds.

Scope: All time · Searchable

A reasoning graph that follows you from Step 1 to residency.

Every concept you touch becomes a node. Every weak topic gets a recovery route. Every session adds a layer. Over time, Stella's memory becomes indistinguishable from a study partner who's been with you since day one.

  • No more "as I mentioned last week" — because she already knows.
  • No more re-explaining patient cases — they're in the L3 ledger.
  • No more forgetting which mechanisms you already nailed — the L2 weak map has it.
  • No more mid-session throttling — token quota scales to your actual study hours.
04
04 — The Compaction Engine

Long session. Permanent memory.
Compresses itself.

When a study session ends, an async Compaction Engine distills the core findings, the diagnostic pitfalls, and the mechanism threads into a permanent entry in your L3 ledger. You never re-cover ground.

Raw session 26 turns · 4,200 tokens
YouWalk me through RTA Type 1 vs Type 4 — I keep confusing the urine anion gap interpretation.
StellaRTA Type 1 has a defect in alpha-intercalated cells, distal H+ secretion fails. So urine pH stays >5.5 even with systemic acidosis. The anion gap math...
YouOK so why does Type 4 present with hyperkalemia but Type 1 doesn't?
StellaType 4 is hypoaldosteronism — aldosterone drives K+ secretion at the principal cell. Without it, K+ builds up. Type 1 is purely a distal H+ issue...
YouHow do you distinguish in a clinical vignette?
StellaLook at serum K+ and urine pH simultaneously. Type 4: hyperkalemic, urine pH <5.5. Type 1: normokalemic, urine pH >5.5...
Compaction
L3 digest · 86 tokens searchable
Concept
RTA Type 1 vs Type 4 — urine anion gap, serum K+ as differentiator
Mechanism
T1: distal H+ secretion defect (α-intercalated). T4: hypoaldosteronism → impaired K+ secretion at principal cell.
Vignette test
Serum K+ + urine pH simultaneously. T1 → normokalemic, pH >5.5. T4 → hyperkalemic, pH <5.5.
L2 status
Needs Work (was Weak prior) · scheduled for re-test in 3 days
05
05 — Clinical engines

Three engines.
One reasoning core.

Engine 01 · Lab Mode

Paste any lab. Get a differential.

Drop any unstructured lab panel into Stella — CBC, LFT, ABG, electrolytes, renal profile, all of it. She identifies abnormals, detects patterns like metabolic acidosis with respiratory compensation, ranks differentials, and gives you exam-ready viva pearls.

Acid-base analysis Pattern detection Ranked differentials Viva pearls Export to Obsidian
Serum K⁺5.9 mmol/L↑ HIGH
Bicarb HCO₃⁻14 mmol/L↓ LOW
Na⁺138 mmol/LNORMAL
Creatinine2.4 mg/dL↑ HIGH
Stella's assessment Pattern suggests hyperkalemic metabolic acidosis — consider RTA Type IV or early AKI. High-yield: calculate anion gap and urine anion gap to differentiate.
Engine 02 · Test Me

Vignettes that teach while testing.

Type /test and Stella generates a 10-question widescreen MCQ block from exactly what you just studied. Every wrong answer triggers a mechanism breakdown — and queues the concept straight into your L2 weak topic tracker for re-test.

USMLE-style vignettes Distractor breakdown Auto-queues L2 L1 / L2 / L3 sync
A 58-year-old man with CHF presents with increasing dyspnoea. Serum sodium is 128 mEq/L. Which best explains the hyponatremia?
A
Nephrogenic diabetes insipidus
B
↑ ADH secretion from reduced effective arterial volume
C
Primary polydipsia
D
Hypothyroidism-induced SIADH
L2 QUEUED Concept auto-tracked as weak topic
Engine 03 · Flashcards

SM-2 spaced repetition, with a twist.

Highlight anything in chat and turn it into a card. Every card supports both flip mode (active text recall) and MCQ mode (quiz-style testing) — switch mid-review with a single tap. Failing a card queues a mechanism-first study note generated by Stella.

SM-2 algorithm Flip ↔ MCQ toggle In-card Q&A Failed-card notes
Concept · Tap to flip
What is the primary enzymatic mechanism behind the negative inotropic effect of β-blockers on cardiac myocytes?
SM-2 · due in 3 days
Mechanism · Tap to flip back
β-blockers antagonize β1 receptors → ↓ cAMP → ↓ PKA activity → ↓ L-type Ca²⁺ channel phosphorylation → less Ca²⁺ influx during the action potential plateau.
L2 · Needs Work
06
06 — Modality

Four ways to engage the same material.

Some moments call for mechanism. Some call for speed. Some call for a USMLE-style throwdown. Switch modes mid-session — Stella keeps the L1 / L2 / L3 context intact while changing how she answers.

🎓
Default · Teach

Structured & mechanistic

Full depth. Headers, mechanisms, traps, pearls. The mode you default to when you actually want to learn something.

💬
Explore

Conversational

Stella picks the shape — paragraph, list, table — based on what fits the question. Best for browsing ideas.

Flash

Razor-sharp

1-6 lines per question. Reasoning off. Pure recall mode for when you're cramming a deck before rounds.

🩺
Vignette

USMLE-style cases

Scenario → A-E. Explanation revealed only after you commit. The exam simulator.

Stella's response mode selector showing Teach, Explore, Flash, and Vignette modes
07
07 — Curriculum

Four exam systems. One reasoning core.

Pick the curriculum. Pick the response mode. Stella routes the entire session through the right clinical engine — no manual configuration, no prompt engineering.

🩺
USMLE

Step 1 & Step 2 CK

First Aid–aligned. NBME-style vignettes. Mechanism-first explanations. Dedicated biostatistics and ethics routing.

📚
MBBS

Indian medical curriculum

Subject-specific modules for anatomy, physiology, biochemistry, pathology, pharmacology, and forensic medicine.

🧬
Mixed

Cross-curriculum mode

Switch seamlessly between USMLE and MBBS framing. For residency-bound students preparing for both ecosystems.

🌿
Chill

Burnout-aware mode

Lower cognitive load. More Socratic. Built for late nights when you can't stare at another wall of text.

08
08 — Orchestration

Your day, synthesized from your actual performance.

The Daily Planner doesn't ask what you want to study — it asks what you should study, based on your L2 weak topics, your last 5 L3 digests, and your manual focus. Outputs an Apple-grade schedule with priority topics, review blocks, MCQ practice, and end-of-day checkpoints.

Stella's Study Workspace showing daily study performance, study analytics, recall performance trend, and recommended medical decks
09
09 — Infrastructure

Three models, routed by what you actually need.

No single model wins every workload. Stella's routing matrix picks the right engine per query — private local, premium clinical, or high-availability cloud failover.

🛡️
Privacy · Speed

Local RTX Routing

Run private, offline, high-speed clinical models directly on your consumer RTX hardware. Your data never leaves your machine. Ideal for sensitive patient data and uninterrupted deep work.

Latency<200ms
🧠
Depth · Clinical

MiniMax Cloud Clinical

Premium clinical reasoning engine optimized for deep pathophysiological analysis, distractor breakdown, and differential ranking. Used for vignettes, lab mode, and mechanism explanations.

Best forVignettes & Labs
☁️
Availability

GLM Cloud Failover

High-speed cloud fail-safe that ensures constant access to study materials even when local servers are offline. Auto-engages when the primary engine is throttled or unavailable.

Uptime99.97%
10
10 — Field reports

From the closed beta.

"I haven't re-explained a patient case in three weeks. My UWorld recall rate went from 60% to 84%."

JM
Jatin M.
MS3 · AIIMS
+24%UWorld
12weeks

"It's the first AI I've used that actually remembers my weak topics across sessions. The Compaction Engine is magic."

AS
Ananya S.
PGY-1 Internal Medicine
3-tiermemory
Dailyactive

"Stella's Lab Mode saved me on a real patient last week. I saw the same pattern in 30 seconds. That's not a chatbot, that's a colleague."

RK
Rohan K.
MS4 · USMLE Step 2
258decks
12kcards
11
11 — Pricing

Pick your intensity.

All plans unlock the 3-tier memory, lab mode, daily planner, flashcards, and the full clinical routing engine. Usage splits into Weekly Usage + a Flex Reserve for bursts — no hard 5-hourly or weekly caps, ever.

Try Out
199
per month
Weekly Usage + Flex Reserve
  • 3-Tier Memory (Lite caps)
  • Lab Mode Interpreter
  • Daily Study Planner
  • Standard clinical routing
Choose Try Out
Basic
499
per month
Weekly Usage + Flex Reserve
  • 3-Tier Memory Architecture
  • Lab Mode Interpreter
  • SM-2 Flashcard Suite
  • Daily Study Planner
  • Standard clinical routing
Choose Basic
Maximum Access
Ultra
1499
per month
Weekly Usage + Flex Reserve
  • Everything in Pro
  • Deepest memory architecture
  • Full model sandbox access
  • Residency-grade prep stack
  • Continuous board training
Choose Ultra
12 — Common questions

Quick answers, no fluff.

Yes. With Local RTX routing enabled, your conversations never leave your machine. Cloud routing is opt-in, and you can switch per-session. We do not train on your data, ever.
With Local RTX routing, yes — fully offline. Cloud models obviously require connectivity, but the routing matrix auto-fails-over to your local engine when the network drops.
Three things they don't have: (1) a 3-tier memory architecture that persists across sessions, (2) exam-specific routing (USMLE, MBBS, Mixed, Chill) with the right persona locked in, and (3) a clinical reasoning engine optimized for medical school — with SM-2 flashcards, lab mode, and a Compaction Engine built in.
USMLE Step 1, USMLE Step 2 CK, MBBS (all years and subjects), and cross-curriculum mode for students preparing for both. We're adding residency-board prep and NEET-PG in Q3.
Yes. No commitment, no contracts. Cancel from your dashboard in two clicks. Your data is exportable to JSON at any time.
Beta users keep their current pricing for life. New cohorts will be priced higher. We're opening access in waves — if you're in, you're locked in.
The Try Out plan (₹199/mo) is the trial — it gives you the full feature set with a lighter usage allowance. Upgrade anytime, downgrade anytime, no questions asked.
Active Beta · 14 slots left this week

Your boards deserve a
real partner.

Stella is in active beta — a closed cohort of medical students stress-testing the memory architecture and clinical engines. New slots open every Monday. No card to start.

Built for USMLE & MBBS · 4.9★ avg · 89% retention