Project atlas
Position Project atlas

Project atlas

240 catalog entries · showing 151–200. 1 published so far.

  • Content Safety Classifier

    Build a transparent moderation aid with category-specific thresholds, appeal/human review, subgroup analysis, and a limitations card.

    • ML-10
    • planned
    • code
    • ai-engineer, backend-engineer
  • Imbalanced Classification and Calibration Lab

    Compare metrics, resampling, class weights, focal loss, Platt scaling, and isotonic calibration on multiple datasets.

    • ML-11
    • planned
    • code
    • ai-engineer, backend-engineer
  • Experiment Analysis Platform

    Perform power and sample-size planning, sequentially safe reporting, guardrail analysis, novelty checks, and segment diagnostics.

    • ML-12
    • planned
    • code
    • ai-engineer, backend-engineer
  • Uplift Modeling Simulator

    Estimate heterogeneous treatment effects on synthetic randomized data and compare treatment policies with honest validation.

    • ML-13
    • planned
    • code
    • ai-engineer, backend-engineer
  • Two-Tower Recommendation System

    Build candidate retrieval, hard-negative sampling, approximate-nearest-neighbor serving, offline metrics, and cold-start strategies.

    • ML-14
    • planned
    • code
    • ai-engineer, backend-engineer
  • Learning-to-Rank Search Stack

    Combine BM25 candidates, behavioral features, pairwise/listwise ranking, offline judgments, and online experiment design.

    • ML-15
    • planned
    • code
    • ai-engineer, backend-engineer
  • AWS Landing Zone in Terraform

    Build modules for IAM, VPC, compute, storage, and RDS with least privilege, remote state, policy checks, and cost estimates.

    • OPS-02
    • planned
    • code
    • ai-engineer, backend-engineer
  • OpenTelemetry Observability Platform

    Correlate metrics, logs, and traces across polyglot services and publish a trace-led incident investigation.

    • OPS-03
    • planned
    • code
    • ai-engineer, backend-engineer
  • SLO and Error-Budget Control Center

    Define user-centered SLIs, multi-window burn alerts, release gates, and a documented 99.9% policy.

    • OPS-04
    • planned
    • code
    • ai-engineer, backend-engineer
  • Incident Command Simulator

    Inject a cascading failure, record decisions and communication, measure detection/mitigation/recovery, and publish a blameless postmortem.

    • OPS-05
    • planned
    • code
    • ai-engineer, backend-engineer
  • Autoscaling Benchmark

    Compare CPU, request, queue-depth, and custom-metric scaling under spikes; quantify cost, stability, and cold-start behavior.

    • OPS-06
    • planned
    • code
    • ai-engineer, backend-engineer
  • Canary Delivery Controller

    Shift traffic progressively, analyze golden signals, pause for approval, and roll back automatically on defined thresholds.

    • OPS-07
    • planned
    • code
    • ai-engineer, backend-engineer
  • Disaster-Recovery Automation

    Restore data and services in a clean region/account simulation, verify integrity, and report achieved RPO/RTO.

    • OPS-08
    • planned
    • code
    • ai-engineer, backend-engineer
  • Cloud Cost Attribution Portal

    Allocate shared costs by service/team/tenant, detect anomalies, recommend savings, and verify recommendations against usage.

    • OPS-09
    • planned
    • code
    • ai-engineer, backend-engineer
  • Secret Rotation Pipeline

    Rotate database and API credentials without downtime, detect stale consumers, revoke old versions, and audit completion.

    • OPS-10
    • planned
    • code
    • ai-engineer, backend-engineer
  • Multi-Region Failover Lab

    Define routing and data-consistency tradeoffs, test regional loss, measure recovery, and document failure modes.

    • OPS-11
    • planned
    • code
    • ai-engineer, backend-engineer
  • Signed Container Supply Chain

    Build minimal images, scan them, create SBOMs, sign artifacts, verify admission, and preserve provenance.

    • OPS-12
    • planned
    • code
    • ai-engineer, backend-engineer
  • Service-Mesh Reliability Lab

    Test mTLS, retries, timeouts, outlier detection, traffic splitting, and the latency/complexity cost of the mesh.

    • OPS-13
    • planned
    • code
    • ai-engineer, backend-engineer
  • eBPF Performance Investigation

    Diagnose a real CPU/network/syscall bottleneck with eBPF tools and validate the fix with repeatable benchmarks.

    • OPS-14
    • planned
    • code
    • ai-engineer, backend-engineer
  • Edge Cache and CDN Simulator

    Implement cache keys, revalidation, stale serving, invalidation, regional latency tests, and origin-protection limits.

    • OPS-15
    • planned
    • code
    • ai-engineer, backend-engineer
  • Codebase Intelligence Assistant

    Index symbols, call graphs, docs, and history; answer with line-level citations, respect repository boundaries, and evaluate maintenance tasks.

    • RAG-02
    • planned
    • code
    • ai-engineer, backend-engineer
  • Incident Knowledge Assistant

    Retrieve runbooks, dashboards, deploys, and past postmortems; surface evidence and uncertainty without taking autonomous production action.

    • RAG-03
    • planned
    • code
    • ai-engineer, backend-engineer
  • Research Synthesis Workspace

    Collect papers and notes, deduplicate claims, cite passages, compare sources, flag conflicts, and export a traceable brief.

    • RAG-04
    • planned
    • code
    • ai-engineer, backend-engineer
  • Multimodal Document RAG

    Retrieve across text, tables, charts, and page images; preserve page coordinates and evaluate answer plus citation correctness.

    • RAG-05
    • planned
    • code
    • ai-engineer, backend-engineer
  • Customer Support Resolution Engine

    Retrieve product and account-safe context, draft grounded responses, abstain on low confidence, and learn from reviewed outcomes.

    • RAG-06
    • planned
    • code
    • ai-engineer, backend-engineer
  • Hybrid Search Reference Engine

    Implement dense retrieval, BM25, reciprocal-rank fusion, filters, reranking, and a labeled relevance benchmark.

    • RAG-07
    • planned
    • code
    • ai-engineer, backend-engineer
  • RAG Evaluation Laboratory

    Version datasets and prompts; measure retrieval recall, citation precision, groundedness, completeness, latency, cost, and regression significance.

    • RAG-08
    • planned
    • code
    • ai-engineer, backend-engineer
  • Adaptive Ingestion Framework

    Parse heterogeneous sources, choose semantic chunk boundaries, deduplicate, detect changes, preserve metadata, and support deletion.

    • RAG-09
    • planned
    • code
    • ai-engineer, backend-engineer
  • Knowledge-Graph Retrieval Study

    Extract entities/relations with provenance, combine graph and vector retrieval, and compare against a strong hybrid baseline.

    • RAG-10
    • planned
    • code
    • ai-engineer, backend-engineer
  • Private Personal Knowledge Vault

    Add local encryption, per-source permissions, export/deletion, transparent memory, and optional offline inference.

    • RAG-11
    • planned
    • code
    • ai-engineer, backend-engineer
  • Multilingual Retrieval System

    Compare multilingual embeddings, translation, language-specific BM25, reranking, and citation quality across languages.

    • RAG-12
    • planned
    • code
    • ai-engineer, backend-engineer
  • Freshness-Aware News and Policy Search

    Track publication/effective dates, superseded sources, incremental indexing, temporal queries, and stale-answer prevention.

    • RAG-13
    • planned
    • code
    • ai-engineer, backend-engineer
  • Air-Gapped RAG Appliance

    Package local ingestion, vector search, reranking, and model inference with resource budgets and no network dependency.

    • RAG-14
    • planned
    • code
    • ai-engineer, backend-engineer
  • Citation and Claim Verifier

    Decompose answers into claims, map each to retrieved evidence, flag unsupported or contradictory statements, and score verifier reliability.

    • RAG-15
    • planned
    • code
    • ai-engineer, backend-engineer
  • URL Shortener at Global Scale

    Design redirect latency, key generation, hot-key handling, analytics, abuse controls, custom domains, and regional failover.

    • SD-01
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer
  • Distributed Rate-Limiting Service

    Design local/global enforcement, algorithm choice, configuration distribution, clock issues, graceful degradation, and tenant fairness.

    • SD-02
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer
  • End-to-End Messaging Platform

    Design delivery, ordering, presence, offline sync, media, encryption boundaries, multi-device state, and abuse response.

    • SD-03
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer
  • Team Collaboration Platform

    Design channels, threads, search, notifications, presence, file sharing, retention, enterprise permissions, and large workspaces.

    • SD-04
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer
  • Cloud File Synchronization

    Design chunking, deduplication, versioning, conflict resolution, metadata, sharing, virus scanning, and regional durability.

    • SD-05
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer
  • Global Video Platform

    Design upload/transcoding, storage tiers, CDN delivery, metadata, recommendations, moderation, and creator analytics.

    • SD-06
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer
  • Real-Time Mobility Dispatch

    Design geospatial indexing, matching, ETA, surge-like load without exploitative pricing, trip state, safety, and city failover.

    • SD-07
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer
  • Git Hosting and Code Review

    Design object storage, repository access, diffs, pull requests, hooks, search, CI events, permissions, and disaster recovery.

    • SD-08
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer
  • Web and Catalog Search Engine

    Design crawling/ingestion, indexing, ranking, freshness, spelling, personalization controls, evaluation, and serving.

    • SD-09
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer
  • Recommendation Platform

    Design event collection, feature pipelines, candidate generation, ranking, exploration, experimentation, feedback loops, and safety.

    • SD-10
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer
  • Vector Search Service

    Design sharding, HNSW/IVF indexes, metadata filters, replication, updates, recall/latency objectives, and multi-tenancy.

    • SD-11
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer
  • AI Agent Platform

    Design tool registry, identity, policy, state, queues, approvals, sandboxing, evaluation, observability, quotas, and failure recovery.

    • SD-12
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer
  • Payment Processing and Ledger System

    Design authorization/capture/refund, double-entry records, idempotency, reconciliation, webhooks, disputes, and auditability.

    • SD-13
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer
  • Social Feed and Timeline

    Design fan-out, ranking, privacy, deletes, celebrity hotspots, caching, freshness, moderation, and experiment hooks.

    • SD-14
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer
  • OWASP Vulnerability-to-Fix Museum

    Build isolated vulnerable examples, exploit them safely, patch them, and add regression tests for major OWASP classes.

    • SEC-02
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer
  • RBAC/ABAC Policy Service

    Evaluate role and attribute policies, explain every decision, test privilege escalation, and cache without stale authorization.

    • SEC-03
    • planned
    • code
    • ai-engineer, backend-engineer, full-stack-engineer