Research Suggestions

Running list of topics, tools, and concepts worth investigating based on knowledge base analysis.

Last Updated: 2026-02-12 (Updated from knowledge edge interview)


High Priority (Gaps in Active Domains)

AI/ML

  • Math → ML bridge - HIGH LEVERAGE. Work through gradient descent by hand using your calculus. Forward pass = matrix multiplication (linear algebra you know) + activation function. Backpropagation = chain rule (you know this). One session transforms your ML understanding. See 2026-02-12-knowledge-edge-map.md.
  • RAG (Retrieval-Augmented Generation) - Pattern for grounding LLMs in your own data. Critical for modern AI applications. You have ML foundations but miss this key pattern.
  • Vector databases - pgvector, Pinecone, Weaviate, Chroma. Storage layer for RAG and semantic search.
  • LLM evaluation frameworks - How to measure if your AI actually works. Look at: RAGAS, LangSmith, Braintrust.
  • Agent architectures - ReAct, tool use patterns, multi-agent systems. Given you’re using Claude Code, understanding this deeply is relevant.
  • ML model provenance via content addressing - Apply SLSA attestation patterns to ML pipelines. Cover training data, code, hyperparameters, environment, AND model weights. Your IPFS + supply chain knowledge is the on-ramp.

Parloa Platform Pre-Reading

Platform Adoption Strategies - How to migrate remaining 60% of services:

  • Conway’s Law and Team Topologies - “Highly aligned, loosely coupled” architectures. Team cognitive load management.
  • Platform as a Product - Internal customer journey, adoption metrics, developer experience optimization.
  • API Design at Scale - Fixing “39 versions in 3 months” problem: versioning strategies, contract-first development, evolutionary architecture.

Conversational AI Platform Engineering:

  • LLM Infrastructure Patterns - Model serving, prompt management, fine-tuning pipelines, A/B testing for conversational AI.
  • Real-time Communication Architecture - WebSocket scaling, session management, conversational state machines.
  • European AI Regulations - EU AI Act compliance for conversational AI platforms, data residency requirements.

Hypergrowth Engineering:

  • Scaling Engineering Teams - Hiring 60 engineers, onboarding at scale, maintaining culture and standards.
  • Distributed System Contracts - Service mesh adoption, API governance, preventing tight coupling during rapid growth.

European Digital Sovereignty

  • GDPR + Cloud Architecture - Data residency, processing locations, vendor dependencies.
  • EU Cloud Services - OVHcloud, Scaleway, Deutsche Telekom alternatives to AWS/GCP/Azure.
  • Digital Markets Act (DMA) - Platform regulation, interoperability requirements.
  • EU Cybersecurity Act - Security certification for digital products, NIS2 Directive compliance.
  • European Startup Ecosystem - Berlin/Munich tech scene, European VC landscape, scaling in EU vs US.

Security

  • Zero Trust Architecture - Notable gap given your security focus. NIST 800-207 is the reference.
  • eBPF for security - Complete. Basic coverage exists.
  • Secret management patterns - HashiCorp Vault patterns, External Secrets Operator, SOPS. You have supply chain but not runtime secrets.
  • Runtime security tooling (Falco, seccomp, AppArmor) - Your security mental model stops at build-time. Falco is the easiest on-ramp given your eBPF awareness.
  • CVE/CWE/CPE taxonomy - Internalize the vulnerability classification system. CVE = specific instance, CWE = class of bug, CPE = affected product. Important given supply chain adjacency.

Platform Engineering

  • Backstage deep-dive - Complete. You have hands-on experience from Thrive Market.
  • Platform Orchestrators - Humanitec, Kratix. Higher-level abstraction than raw K8s.

Medium Priority (Natural Extensions)

DevOps/SRE

  • Kingman’s formula & queueing theory for engineers - HIGH LEVERAGE. Connects Little’s Law (which you know) to back-pressure in distributed systems. The math: wait ∝ 1/(1-ρ). At 90% utilization, wait = 9x processing time. Bridges your theory and systems practice.
  • Back-pressure patterns - Buffering/queuing, rate limiting, load shedding, pull-based consumption. You know the reactive side (circuit breakers) but not the proactive side. This IS queueing theory applied to distributed systems.
  • OpenTofu - Terraform fork. Given your GitOps focus, worth tracking.
  • Dagger - CI/CD pipelines as code. Interesting alternative to YAML-based CI.
  • FinOps - Cloud cost management. Your notes don’t touch costs.
  • Cilium - eBPF-based networking. Connects your K8s and security interests.
  • Alert calibration as precision/recall - Frame alert-to-incident ratio as precision, incident-without-alert rate as recall. Connects your stats knowledge to SRE practice.

Observability

  • OpenTelemetry Collector patterns - You have OTel basics, but collector pipelines are where it gets interesting.
  • Continuous Profiling - Parca, Pyroscope. The “fourth pillar” of observability.
  • ML Observability - MLflow, Weights & Biases, Arize. Bridge your ML and observability knowledge.

Statistics

  • Bayesian A/B testing - Your stats are frequentist. Bayesian approach is increasingly preferred for experiments.
  • Causal inference - Difference-in-differences, instrumental variables. Beyond correlation.
  • Sequential testing - When to stop experiments early. Connects to progressive delivery.

Lower Priority (Interesting Adjacencies)

P2P/Decentralized

  • Farcaster - Newer social protocol gaining traction. Different architecture than Bluesky/AT Protocol.
  • Ceramic Network - Decentralized data streams. Relevant to your Noosphere interest.
  • Local-first software hard problems - Beyond CRDTs/sync: schema evolution without coordinated migrations, access revocation of replicated data, CRDT tombstone accumulation/GC. You have good product instincts here (network effects, telemetry); the technical walls are different.

Programming

  • WebAssembly + WASI - Portable compute. Connects to edge computing, P2P, and platform engineering.
  • Nix for development environments - You have “Rust on Nix” but Nix as a dev tool is broader.

Architecture

  • Cell-based architecture - Isolation pattern for resilience. Used at AWS, Slack.
  • Hexagonal architecture - Ports and adapters. Clean separation of concerns.
  • Event-Driven Architecture Patterns - Relevant for conversational AI state management and real-time systems.

Books Worth Reading

Based on your interests and Parloa context:

  1. “Designing Data-Intensive Applications” (Kleppmann) - REREAD chapters 7-9 specifically. Your CAP understanding is basic; Kleppmann’s PACELC framing and consistency model taxonomy would fill the gap.
  2. “Team Topologies” (Skelton & Pais) - Complete
  3. “Building Secure & Reliable Systems” (Google SRE) - Bridges your security and SRE interests
  4. “Staff Engineer” (Larson) - Given your seniority, relevant for career
  5. “Thinking in Systems” (Meadows) - Complete
  6. “Platform Engineering on Kubernetes” (Krief) - Practical patterns for K8s-based platforms
  7. “The Technology Fallacy” (Kane et al.) - Digital transformation and platform adoption strategies
  8. “Conversational AI” (McTear et al.) - Technical foundations for Parloa’s domain

Tools to Try

Done?ToolWhyEffort
DONEOllamaRun LLMs locally, useful for RAG experimentsLow
pgvectorVector search in Postgres, minimal new infraLow
REJECTEDBackstageSpin up locally to understand deeplyMedium
FalcoRuntime security, see what it catchesMedium
DaggerAlternative CI/CD approachMedium

Conference Talks to Watch

  • KubeCon Platform Engineering Day recordings - State of the practice
  • GOTO conferences - Often have good system design content
  • Strange Loop (archived) - Deep technical content
  • Papers We Love meetup recordings - Academic papers made accessible

Questions to Explore

These emerged from analyzing your notes and Parloa context:

  1. How does “credible exit” (from Bluesky) apply to enterprise platform engineering?
  2. Can supply chain security practices apply to ML model provenance?
  3. How do DORA metrics apply (or not) to ML/AI teams?
  4. What platform patterns enable “highly aligned, loosely coupled” at hypergrowth scale?
  5. How do you prevent API sprawl (39 versions in 3 months) while enabling team autonomy?
  6. What does European digital sovereignty mean for conversational AI platforms?
  7. How do platform adoption metrics differ for AI/ML workloads vs traditional services?

How This List is Maintained

  • Added during vault analysis sessions
  • Checked off when notes exist covering the topic
  • Re-prioritized quarterly based on relevance
  • Items older than 1 year without action get pruned

Maintained by Claude. Last comprehensive review: 2026-02-12 (knowledge edge interview)