AgenticBricks is an AI consulting firm that embeds expert engineers directly with enterprise clients to tackle complex, high-stakes challenges. We specialize in agentic systems, intelligent automation, and LLM-powered solutions that create measurable outcomes.
Applied Scientist
Type: Full-time
The role
We are hiring an Applied Scientist at AgenticBricks supporting a large ecommerce retailer. You'll own the core ML lifecycle end to end — feature engineering, building and training models in production, and inference at scale — for systems that serve real customers and operations at major online retail volume. This is hands-on applied work: your output is production models and the pipelines that feed and serve them, not prototypes that stop at a notebook.
What you'll do
Feature engineering
- Design, build, and maintain features from large-scale, messy retail data — transactions, catalog, behavioral signals, supply-chain and operational data.
- Build reliable feature pipelines (batch and streaming) and the transformations behind them, with an eye toward correctness, freshness, and reuse across models.
- Work with feature stores and data infrastructure so the same features are consistent between training and serving, and debug train/serve skew when it shows up.
Models built in production
- Train, validate, and productionize models against real production data and infrastructure — not sandboxed datasets.
- Stand up reproducible training pipelines: versioned data and features, automated retraining, and the evaluation gates that decide what ships.
- Tune for the realities of scale and cost, and design the offline and online experiments (including A/B tests) that prove a model is actually better.
Inference
- Build and optimize model serving for production — batch, real-time, and low-latency online inference under retail traffic loads.
- Own the operational side of inference: latency, throughput, cost, autoscaling, monitoring, and drift detection.
- Diagnose and fix production model issues quickly, and close the loop between what's observed in serving and what gets fixed in features or training.
What we're looking for
- Graduate degree in a quantitative field (ML, CS, statistics, applied math) or equivalent applied experience.
- Strong ML fundamentals plus the statistical literacy to evaluate models honestly.
- Strong programming skills (Python and the standard ML/data stack) and the ability to write production-quality code other engineers build on.
- Demonstrated experience taking models all the way to production — feature pipelines, training, and serving — at meaningful scale.
- Practical command of the full pipeline: feature engineering, reproducible training, and inference/serving, including the failure modes at each stage.
- Clear communication; you can explain a method, a result, and a caveat to a non-technical stakeholder.
Nice to have
- Experience in ecommerce, retail, marketplace, or large-scale consumer products.
- Hands-on work with feature stores, streaming pipelines, distributed training, or model-serving infrastructure.
- Familiarity with recommendation, search/ranking, or forecasting systems.
- MLOps depth: CI/CD for models, monitoring, versioning, and drift management in production.