Senior Software / ML Engineer — Dynamic Pricing Engine

We are looking for a seasoned software engineer with a decent machine-learning background to design, build, and ship the dynamic pricing engine for a live sports-wagering product that we are developing at the moment. This is a long-term project that we are working on for one of our international customers. The role is full-stack: you will own everything from real-time event ingestion to probabilistic modelling, odds optimization, and the dashboards that let both us and the client see what the model is doing.

The product:
We are building a monitoring and wagering platform for golf courses: players can bet on the outcome of their own shots, and the system offers live payout ratios for the next shot. Your job is the engine that prices those shots — ingesting shot outcomes as they happen, updating the underlying probability model online, and producing payout ratios that maximize revenue under defined risk and return-to-player constraints.

It's a genuinely interesting modelling problem, because the bettor is the player — which means the engine has to price defensively against private information and self-influenced outcomes, not just estimate a success rate. We'll expect you to engage with that, not paper over it.

Responsibilities
* Design and build the streaming ingestion pipeline for shot-level events (player, lie, distance, club, conditions, outcome) on a real-time message bus.
* Build and maintain the probability model that serves calibrated outcome probabilities for the next shot — hierarchical/Bayesian core with online (e.g. Beta-Binomial / Dirichlet-Multinomial) updates, plus richer context models where they earn their place.
* Implement probability calibration and continuous calibration monitoring (calibration drift is a direct cash leak — you'll treat it as a first-class metric).
* Build the pricing layer: converting probabilities into offered odds with a principled margin model (e.g. Shin-style margins for markets containing informed money).
* Build the optimization layer that tunes margins against a demand/elasticity model to maximize revenue under liability and return-to-player constraints (contextual bandits / Thompson sampling).
* Build the risk & exposure engine: live liability tracking, stake caps, per-player odds shading, and anomaly detection for abuse patterns.
* Build visualization and monitoring — dashboards for model state, P&L, calibration, and exposure that are legible to non-technical stakeholders.
* Work directly and frequently with the client (English-speaking) on technical direction, trade-offs, and results.

Requirements
* IMPORTANT: Fluent, confident spoken and written English. Our client is an English-speaking customer (Australian) who is keenly interested in the technical detail and will interact with you regularly. You must be able to discuss modelling decisions live, defend trade-offs, and explain complex statistical ideas to a non-technical audience clearly and patiently.
* 8+ years building production software, with strong Python and solid software-engineering fundamentals (testing, CI, observability, clean service design). You ship things that run, not just notebooks.
* Strong applied ML / statistics background — probabilistic modelling, Bayesian inference, online/streaming updates, and an instinct for calibration and uncertainty. You understand why a model is wrong, not just that it is.
* Hands-on experience with real-time / streaming data systems (Kafka, Kinesis, or similar) and the operational realities of low-latency serving.
* Experience taking an ML system end to end — ingestion, training/updating, serving, monitoring, retraining — i.e. MLOps in practice, not just model fitting.
* Comfort building the visualization and monitoring surface yourself (dashboards, metrics, plots) rather than waiting for someone else to.
* Sound judgement about revenue, risk, and constraints — you can think about an objective function that isn't just "make the number go up."

Nice to have
Background in sports betting, market-making, quantitative trading, pricing, or actuarial work — any domain where you've priced uncertain outcomes against informed counterparties.
Familiarity with probabilistic programming frameworks (PyMC, Stan, NumPyro) and/or gradient-boosted models with calibration.
Experience with bandits / reinforcement learning for online decision-making.
Experience with fraud / abuse / anomaly detection on user-level behavioural data.
Awareness of the regulatory and responsible-gambling dimension of wagering products, and willingness to build the necessary hooks in from day one.
Dashboarding / BI tooling (Grafana, Plotly/Dash, Streamlit, or similar).

A bit about us
We are a full-cycle engineering company. We build products from idea to mass production for customers in America and Europe. We have development offices in Moscow and Ivanovo, we manufacture electronics in Taiwan, and we develop software in Vietnam. We take on hard projects, we move fast, and we have genuinely level-headed colleagues.

A few recent and currently active projects:
* Golf-course monitoring system. Custom panoramic cameras on RK3588 with video streaming over LTE, radar integration for ball tracking, and advanced software to manage a fleet of devices across many courses — plus the wagering and pricing platform this role is for. A very large and complex full-stack project (electronics, firmware, UX/UI, back/front, mobile, third-party integration).
* AI-based voice assistant in a flat cradle form factor that docks an iPhone over DP Alt Mode, converts screen frames to JPEG on an FPGA, ships them to the cloud, and replays the returned mouse/keyboard actions back to the phone — all on a 9×53 mm board.
* Next-generation e-bike family with digital steer-by-wire control: full mechanical, electronic, and firmware design, four processing nodes, an onboard CAN bus, custom bootloaders, and cascaded firmware updates.
* Smartwatch with onboard maps (i.MX RT500), a competitor to the Garmin Fenix — full electronics, mechanical, firmware, and cloud development across four countries.
Interactive basketball backboard (NVIDIA Orin / RK3588): ML pipelines for court segmentation, player and ball recognition, ball-trajectory estimation, and make/miss detection.

We support our customers at every step of the product development journey - from PoC and pre-EVT prototypes to mass production in China and Taiwan. If you enjoy new technology and like building sophisticated products for an international market, we'd love to talk.

Conditions
* Full-time on our projects. Long-term collaboration prospects.
* Remote, or in one of our offices.
* Competitive salary; paid vacation per the labour code .

If you are interested, please reply to this job post in English and share with us a bit of your background. Only replies in English will be considered.