15 set - Roma
Neurons Lab
ph3About The Project (description, Duration, Stage) /h3 pHands-on bTech Lead /b for an bAI Companion /b in an online mahjong game. The client is a bsocial gaming company (web3 element) /b that scales its product and team. We deliver the AI side of their game as their embedded AI partner. The AI Companion plays mahjong at a strong level and explains its moves. The core of the role is to bbuild the mahjong-playing algorithm /b: a dedicated decision-making model (RL, imitation learning, or search-based — trained on the client's hand-history data) with an bLLM reasoning layer /b on top. Key design constraints: a bvalid-action contract /b with the game engine (the bridge supplies legal moves), bwin detection /b, and a b2-second response budget /b per move. Explanations run async. Support for more than one rule set (riichi and regional variants) is on the roadmap. /p pbDuration /b: 3 months, 0.5 FTE. /p h3What You’ll Actually Do (example Tasks) /h3 ul liDesign and build the mahjong-playing algorithm: choose and defend the approach (imitation learning on hand histories, RL / self-play, search with MCTS, or a hybrid), then train, evaluate, and ship it. /li liOwn the technical architecture end to end: game model + LLM reasoning layer, valid-action mask, win detection, and the API contract with the client's game bridge. /li liHit the 2-second response budget: design and measure the inference path, batching, and caching; keep a latency buffer for the client-facing number. /li liDefine what data and event names we need from the client (hand histories, event streams); build the training and calibration pipeline on that data. /li liBuild and run the evaluation harness: measure play strength against the client's reference points, and validate explanation quality.
/li liStand up LLM observability with Langfuse (async logging, N+1 batch) as an early sprint quick win. /li liTake over context from Vlad Borysenko (0.15–0.2 FTE supervision during ramp-up) and lead the sprint work with the AI Engineer; work with the client's Product Owner in a scrum process. /li liFront the client's CTO and engineers on technical decisions; explain trade-offs in plain language and in depth when asked. /li liWatch the risks the account team flagged: licensing on new training data, engine-bridge capabilities, and multi-rule-set scope. /li /ul h3Skills (hands-on first) /h3 ul liGame AI / sequential decision-making: hands-on RL, imitation learning, or search-based agents (MCTS, self-play) — ideally for imperfect-information games (mahjong, poker, card games) /li liExpert Python for ML systems; strong software engineering (APIs, testing, CI) /li liModel training on gameplay data end to end: data → training → evaluation → serving /li liLLM application engineering: reasoning layers, prompt and context design, structured outputs, guardrails /li liLow-latency inference: profiling, batching, caching, model-size trade-offs against a hard time budget /li liLLM observability and evaluation (Langfuse or similar)
/li liAWS deployment for ML workloads /li liTechnical leadership of a small pod; clear written and spoken communication with client engineers and executives /li /ul h3Knowledge /h3 ul liGame theory for imperfect-information games; evaluation of play strength (win rates, Elo-style ratings, baseline agents) /li liGame-engine integration patterns (event streams, action masks, state bridges) /li liWeb3 / gaming product context — plus, not required /li liAWS Well-Architected for ML workloads /li /ul h3Experience /h3 pKey characteristics (ideally 4/4): /p ul liHands-on ML/AI engineering at production scale /li liShipped an AI system inside a live product with hard latency limits /li liCloud hyperscaler experience (AWS preferred) /li liTechnology consulting / client-facing delivery background /li /ul pRole-specific characteristics: /p ul li6+ years hands‑on ML/AI engineering, with real game AI or sequential decision-making work (RL / MCTS / self-play — not only LLM apps) /li liTrained models on user or gameplay data end-to-end (data → training → evaluation → serving) /li liLed small delivery teams while still coding personally /li liComfortable owning an architecture in front of a technical client CTO /li /ul h3Questions for Applicants /h3 ul liImperfect information: mahjong hides most tiles from each player. How does hidden information change your algorithm choice compared to a perfect-information game like chess? /li liLatency budget: tell us about a system you shipped with a hard response-time limit. How did you design, measure, and defend the budget? /li liLLM + model hybrid: how would you combine a trained game model with an LLM explanation layer so the explanation never contradicts the move? /li liHands‑on + lead: how do you balance personally coding the hard parts with leading an engineer and fronting the client? /li /ul /p #J-18808-Ljbffr
16 set - Corsico
Cantine
16 set - Cascina
Technoform
16 set - Balgano
Concentrix
16 set - Torino
GRANDE FOREDIT