🔬 Applied Scientist & ML Specialist938 Problems · 127 Topics🌐 Linked to Tech Vault

TalentMe Science Depth: Universal Applied Science & ML Depth Problem Bank

938 end-to-end science depth, system design, transition bridge & BQ problems covering derivations, production trade-offs, pitfalls, and follow-ups.

🎯

Science Depth Mastery Tracker

0% Mastered

Locally synced in your browser. Track, review, and master all 938 scientific & design topics.

Mastered: 0
Reviewing: 0
Unstarted: 25
🚀 Start Next Unpracticed →
🌐
Linked to TalentMe Tech Vault & MindmapsM9: Senior ML & AI System Design
All 938 questions are directly linked to TalentMe's Tech Vault guides, atomic cards, and interactive mindmaps.
🧭ML & AI System Design Mindmap→📖4-Stage RecSys & Enterprise AI Design→🎯16-Dim Registry Hub→
Found 24 in-depth questions
M9-001Senior ML & AI System DesignRecSys, Ads & Search Architecture
Hard

Design a Billion-Scale Four-Stage Feed Recommendation Architecture

⚡ One-Sentence Answer

Balance candidate scale and compute complexity across four stages: Dual-Tower ANN recall, distilled pre-ranker, MMoE fine ranker, and DPP diversity re-ranking within 100ms P99.

Math & Tradeoffs→
M9-002Senior ML & AI System DesignRecSys, Ads & Search Architecture
Hard

Design a Real-Time Ads Ranking, eCPM Calibration & Budget Pacing Architecture

⚡ One-Sentence Answer

Enforce auction integrity via well-calibrated eCPM bids, negative-downsampling log-odds correction, delayed conversion survival reweighting, and closed-loop PID throttling.

Math & Tradeoffs→
M9-003Senior ML & AI System DesignRecSys, Ads & Search Architecture
Hard

Design a Cold-Start Item & Creator Exploration Pipeline

⚡ One-Sentence Answer

Break zero-exposure cold start using multimodal priors trained with random ID-dropout, tiered seed pools, contextual bandits, and IPS position debiasing.

Math & Tradeoffs→
M9-004Senior ML & AI System DesignRecSys, Ads & Search Architecture
Medium

Design an E-Commerce Semantic + Lexical Product Search & Re-Ranking Engine

⚡ One-Sentence Answer

Combine NER slot pre-filtering (zero wrong-SKU errors) with BM25+Vector RRF hybrid retrieval and relevance-gated GMV ranking.

Math & Tradeoffs→
M9-005Senior ML & AI System DesignRecSys, Ads & Search Architecture
Medium

Design a Notification Timing, Volume & Unsubscribe-Guard System

⚡ One-Sentence Answer

Target persuadable users via Uplift Modeling and solve daily send-volume vs permanent opt-out trade-offs using Lagrangian multipliers.

Math & Tradeoffs→
M9-006Senior ML & AI System DesignRisk, Pricing & Large-Scale Retrieval
Hard

Design a Sub-10ms Real-Time Payment Fraud Detection System

⚡ One-Sentence Answer

Meet <=10ms P99 SLAs by pushing multi-hop GNN risk into async Redis updates while running 1ms rules, sliding Flink windows, and in-process ONNX GBDTs synchronously.

Math & Tradeoffs→
M9-007Senior ML & AI System DesignRisk, Pricing & Large-Scale Retrieval
Hard

Design a Real-Time Dynamic Pricing & Supply-Demand Elasticity Architecture

⚡ One-Sentence Answer

Isolate true conditional price elasticity via Double ML with monotonic constraints, optimize bounded multipliers, and run time-switchback experiments.

Math & Tradeoffs→
M9-008Senior ML & AI System DesignRisk, Pricing & Large-Scale Retrieval
Medium

Design a Trillion-Token Pretraining Corpus Deduplication & Decontamination Pipeline

⚡ One-Sentence Answer

Scale corpus cleaning to trillions of tokens via cheap heuristic sieves, b-band MinHash-LSH fuzzy dedup, quality classifiers, and 13-gram eval decontamination.

Math & Tradeoffs→
M9-009Senior ML & AI System DesignRisk, Pricing & Large-Scale Retrieval
Medium

Design a Sub-15ms Personalized Search Typeahead Autocomplete Architecture

⚡ One-Sentence Answer

Deliver <=15ms autocomplete suggestions via compressed memory-resident FSTs storing precomputed top-K candidates per prefix, paired with atomic mmap swaps and safety blocklists.

Math & Tradeoffs→
M9-010Senior ML & AI System DesignRisk, Pricing & Large-Scale Retrieval
Medium

Design a Multimodal Document & Image Search Engine (SigLIP + ColPali MaxSim)

⚡ One-Sentence Answer

Avoid chart/table detail collapse in single-vector embeddings by pairing fast binary-quantized recall with patch-level ColPali MaxSim late interaction.

Math & Tradeoffs→
M9-011Senior ML & AI System DesignLLM Serving, RAG & Agent SysDesign
Hard

Design a 100K-QPS LLM Gateway (PD Disaggregation, Prefix Routing & Semantic Cache)

⚡ One-Sentence Answer

Eliminate prefill-vs-decode latency spikes via RDMA-connected disaggregated worker pools, consistent prefix-hash routing, semantic caches, and cascading routers.

Math & Tradeoffs→
M9-012Senior ML & AI System DesignLLM Serving, RAG & Agent SysDesign
Hard

Design a Zero-Leak Multi-Tenant RBAC Enterprise Hybrid RAG Hub

⚡ One-Sentence Answer

Guarantee strict enterprise access isolation inside retrieval via filterable HNSW traversal paired with Docling layout parsing and idempotent CDC indexing.

Math & Tradeoffs→
M9-013Senior ML & AI System DesignLLM Serving, RAG & Agent SysDesign
Hard

Design a Durable Checkpointable & Idempotent Multi-Agent Execution Harness

⚡ One-Sentence Answer

Move beyond fragile memory ReAct loops using Postgres-checkpointed state machines with step/budget breakers, idempotent tool keys, HITL pause interrupts, and subagent isolation.

Math & Tradeoffs→
M9-014Senior ML & AI System DesignLLM Serving, RAG & Agent SysDesign
Hard

Design a Sub-Second Ephemeral MicroVM Sandbox Platform for AI Coding Agents

⚡ One-Sentence Answer

Execute untrusted AI-generated code safely using Firecracker MicroVM warm snapshots (<150ms boot), CoW overlay caches, and strict zero-trust network egress.

Math & Tradeoffs→
M9-015Senior ML & AI System DesignLLM Serving, RAG & Agent SysDesign
Hard

Design a Full-Duplex Sub-400ms Voice-to-Voice Agent with Instant Barge-In

⚡ One-Sentence Answer

Build natural sub-400ms voice agents over WebRTC with semantic turn detection, 30ms echo-canceled barge-in truncation, and clause-level streaming TTS synthesis.

Math & Tradeoffs→
M9-016Senior ML & AI System DesignFeature Store, Sharding & CI Infra
Hard

Design a Zero-Skew Online/Offline Feature Store with Point-in-Time Travel

⚡ One-Sentence Answer

Guarantee zero training-serving skew via shared streaming feature transforms feeding Redis online stores and Iceberg offline tables joined strictly via AS-OF time travel.

Math & Tradeoffs→
M9-017Senior ML & AI System DesignFeature Store, Sharding & CI Infra
Medium

Design a 10-Billion-Scale Distributed Vector Search Architecture

⚡ One-Sentence Answer

Scale vector retrieval to 10B records using in-memory quantized sketches + NVMe FP16 rescoring (DiskANN/IVF-PQ), tenant shard routing, and background LSM compaction.

Math & Tradeoffs→
M9-018Senior ML & AI System DesignFeature Store, Sharding & CI Infra
Medium

Design a Layout-Aware Document ETL & Parent-Child RAG Indexing Pipeline

⚡ One-Sentence Answer

Cure chunk-boundary collapse with layout-aware parsing (Docling), header breadcrumb prefixes, sharp 200-token child index vectors, and rich 1,500-token parent expansions.

Math & Tradeoffs→
M9-019Senior ML & AI System DesignFeature Store, Sharding & CI Infra
Medium

Design a Shadow-Traffic & Canary Auto-Rollback Release Gate for LLMs

⚡ One-Sentence Answer

Safely roll out prompt/model revisions through offline golden CI assertions, async non-blocking shadow differential audits, and fast metric-driven canary rollbacks.

Math & Tradeoffs→
M9-020Senior ML & AI System DesignFeature Store, Sharding & CI Infra
Medium

Design a Privacy-Aware Edge-Cloud Hybrid AI Inference Architecture

⚡ One-Sentence Answer

Route local-sensitive and routine tasks onto zero-cost on-device NPUs while escalating complex masked context cleanly to frontier cloud APIs based on difficulty & thermals.

Math & Tradeoffs→
M9-021Senior ML & AI System DesignFeature Store, Sharding & CI Infra
Hard

Design a Large-Scale RLHF & GRPO Post-Training Distributed Platform

⚡ One-Sentence Answer

Solve post-training's rollout bottleneck using Hybrid GPU engines (vLLM rollouts + FSDP/Megatron trainers), CPU sandbox verifiers, and async straggler cutoffs.

Math & Tradeoffs→
M9-022Senior ML & AI System DesignFeature Store, Sharding & CI Infra
Medium

Design a Closed-Loop Production-Trace-to-Golden-Dataset Eval Flywheel

⚡ One-Sentence Answer

Continuously harvest anonymized production failures into versioned golden test suites guarded by position-debiased LLM judges calibrated to human Cohen's Kappa >= 0.75.

Math & Tradeoffs→
M9-023Senior ML & AI System DesignFeature Store, Sharding & CI Infra
Hard

Design a Zero-Trust AI Security Proxy (Reversible PII Tokenization & Parallel Guardrails)

⚡ One-Sentence Answer

Inspect every AI egress call through reversible session-scoped PII tokenization, parallel injection classifiers (<=35ms overhead), TPM quota enforcement, and WORM audit trails.

Math & Tradeoffs→
M9-024Senior ML & AI System DesignFeature Store, Sharding & CI Infra
Medium

Design a Multi-Month Long-Horizon Agent Memory System

⚡ One-Sentence Answer

Structure multi-month memory cleanly across working sliding rings, time-decayed episodic vectors, and bi-temporal structured user profiles updated out-of-band.

Math & Tradeoffs→