codax
Founding AI / ML Engineer ( Personalisation )
codax, San Francisco, California, United States, 94199
Overview
Founding AI / ML Engineer (Personalisation)
Location: SF Bay Area (hybrid) or In-Person (Preferred)
Office: 505 Howard Street
Vision
We’re building the reasoning layer for customer experience, a privacy-first, explainable system that understands why each person buys, safely simulates outcomes, and orchestrates touch points in real time at near-zero marginal cost. It plugs into any stack as a vendor-agnostic brain and compounds into an autonomous orchestrator that maximises LTV and margin.
We already have distribution in motion via enterprise CX Partners, top-agency advisors (WPP, OMC), and committed angels.
Responsibilities
Own the ranking / reco core: features, training, eval, online inference.
Design cost-aware intelligence (keep the “why” without an LLM per event).
Build offline generation + fast online serving / caching; handle cold-start / backfills.
Ship with design partners; instrument and prove lift quickly.
Qualifications
Contextual data capture & featurisation (surveys / free text → embeddings / NLP).
Cost-aware pre-computation; avoid per-event LLM; cadence-based refresh.
Customer clustering & query taxonomy; catalog-aware resolution limits.
Offline / online architecture; fast (profile / query) lookups; cache coherency.
Cold-start strategies, backfills, and refresh triggers on new cohorts / catalog updates.
Content personalisation for key discovery / conversion surfaces; multi-channel outputs.
Profile update logic after actions; consistent state across caches / stores.
3–5 yrs ML with shipped impact in personalisation / CX (Klaviyo / Braze / Dynamic Yield / Uber / DoorDash-like).
Stack: Python, PyTorch, SQL; Spark / Beam, Airflow, Kafka / Kinesis, Redis; GCP / AWS. Feast / Tecton a plus.
Grit: bias to ship, founder hours, comfort with ambiguity.
Nice to have
E-commerce / subscription, privacy-by-design, RL / bandits / causal, simulation / synthetic personas.
Comp & setup Meaningful founding equity + salary. Part-time → full-time post-raise or full-time now.
#J-18808-Ljbffr
Location: SF Bay Area (hybrid) or In-Person (Preferred)
Office: 505 Howard Street
Vision
We’re building the reasoning layer for customer experience, a privacy-first, explainable system that understands why each person buys, safely simulates outcomes, and orchestrates touch points in real time at near-zero marginal cost. It plugs into any stack as a vendor-agnostic brain and compounds into an autonomous orchestrator that maximises LTV and margin.
We already have distribution in motion via enterprise CX Partners, top-agency advisors (WPP, OMC), and committed angels.
Responsibilities
Own the ranking / reco core: features, training, eval, online inference.
Design cost-aware intelligence (keep the “why” without an LLM per event).
Build offline generation + fast online serving / caching; handle cold-start / backfills.
Ship with design partners; instrument and prove lift quickly.
Qualifications
Contextual data capture & featurisation (surveys / free text → embeddings / NLP).
Cost-aware pre-computation; avoid per-event LLM; cadence-based refresh.
Customer clustering & query taxonomy; catalog-aware resolution limits.
Offline / online architecture; fast (profile / query) lookups; cache coherency.
Cold-start strategies, backfills, and refresh triggers on new cohorts / catalog updates.
Content personalisation for key discovery / conversion surfaces; multi-channel outputs.
Profile update logic after actions; consistent state across caches / stores.
3–5 yrs ML with shipped impact in personalisation / CX (Klaviyo / Braze / Dynamic Yield / Uber / DoorDash-like).
Stack: Python, PyTorch, SQL; Spark / Beam, Airflow, Kafka / Kinesis, Redis; GCP / AWS. Feast / Tecton a plus.
Grit: bias to ship, founder hours, comfort with ambiguity.
Nice to have
E-commerce / subscription, privacy-by-design, RL / bandits / causal, simulation / synthetic personas.
Comp & setup Meaningful founding equity + salary. Part-time → full-time post-raise or full-time now.
#J-18808-Ljbffr