Work with me

Adam Rybinski

Who I work with & what I do

I work with startups building agentic features, data/ML teams who need their LLM outputs to be provably consistent, and game studios with complex economy rules that can't tolerate bugs. If your team ships logic that needs to be auditable, contradiction-free, and evolvable — you're my client.

Typical engagements: AI product startups, in-house ML platform teams, game economy designers, fintech compliance rule authors, and R&D groups exploring neuro-symbolic stacks.


Problems I solve


Ways we can work together

LLM + ASP architecture design

Blueprint for a neuro-symbolic stack. We map your domain, define the solver constraints, wire the LLM proposal loop, and ship a working integration with evaluation harness. 2–4 weeks.

Knowledge engineering sprints

I encode your domain rules in ASP or Prolog, build a testable knowledge base, and set up the gap-analysis pipeline so the system tells you what it's missing. 1–3 weeks per sprint.

Agentic workflow design

Multi-model orchestration with safety rails, evaluation harnesses, and fallback logic. I design the routing protocol, the verification gates, and the human-in-the-loop review layer. 2–4 weeks.

Game economy / market logic audits

I formalise your economy rules, market dynamics, and exploit surfaces as a logical model, then run property-based tests and solver-based verification to find contradictions before they cause problems. 1–3 weeks.


How engagements work

  1. 30–45 min call to map your domain, risks, and constraints. No pitch.
  2. Short written proposal with scope, timeline, and fixed price.
  3. Iterative delivery with testable artifacts — rule files, CLIs, evaluation harnesses. You see progress every few days.
  4. Handover + docs plus optional ongoing advisory (monthly retainer).

Selected work & proofs

Built ASP-based knowledge verification layer LiveKnowledge: closed-loop knowledge engineering with LLMs + Clingo. Open source. GitHub · Architecture overview
Designed multi-agent workflow for RAG + symbolic reasoning Client: fintech compliance team. Reduced false-positive rule triggers by ~60% through formal verification of extracted facts.
Audited game economy rule set Client: mobile game studio. Found 12 latent contradictions in the published economy model. Produced a test suite that catches regressions on every patch.

Stack & focus areas

Clingo / ASP Prolog ILP / Meta-interpretive learning OpenAI HuggingFace Multi-agent workflows Evaluation harnesses Python TypeScript Docker RAG + symbolic layers LiveKit / real-time voice

Fit & non-fit

I take 1–2 concurrent clients. Minimum engagement is 2 weeks.

Good fit: LLM features that need verification, domain rule formalisation, agentic architecture with safety requirements, game/market logic that matters.

Not a good fit: generic CRUD applications, pure front-end work, unpaid trials or speculative pitches, roll-your-own chatbot that just needs a prompt template.

Pricing: fixed-fee packages and monthly retainers. Most projects start around $5k–15k depending on scope.


Ready to talk?

If you need LLM features that behave like verified systems rather than hopeful heuristics, tell me about your stack.

Email: [email protected]

LinkedIn: linkedin.com/in/adam-rybinski

Request a consultation →