3+ years of experience building and deploying Machine Learning models in production.
1+ years of experience with VLM/LLM based features in production.
Proficient in Python backend frameworks like FastAPI or Django.
Strong relational data modeling skills in Postgres, including schema design and migrations.
Experience with high-volume data pipelines and agentic workflows.
Familiarity with asynchronous work patterns and AWS services.
Ability to write tests and debug production issues effectively.
Responsibilities
Own features from design through release and continuous improvement.
Collaborate with Full Stack engineers to integrate models into production features.
Partner with product and design teams to identify key problems to solve.
Optimize model performance for accuracy, latency, and reliability.
Enhance architecture, testing, and observability practices within the team.
Contribute to shaping the product and technical roadmap.
Benefits
Real stakes with actual customers, impacting revenue directly.
Opportunity to work on a product with existing paying customers and proven outcomes.
Chance to define a new category in agentic software for physical infrastructure.
Join a small, senior team with a flat structure that encourages open debate and quick decision-making.
Full Job Description
The role
We're looking for an Applied AI Engineer to help build the core ML engine of the Foresight platform.
You will build the systems that enhance the ability for efficient delivery of large scale construction projects. You will architect, train and deploy the models that perform Foresight's risk predictions and AI recommendations, to be used in interfaces that make 100k-activity projects understandable at a glance. This is a hands-on role on a new team, for someone who owns models and backend features from training to deployment and monitoring to build systems that are accurate, scalable and observable.
At Foresight, machine learning is an integral part of the platform. You will be using complex, multimodal high stakes data to make predictions that will be used by teams to deliver major projects on time. This could translate to work on:
Risk prediction: Build, train and deploy models that predict if a project will be at risk and by how much.
Criticality prediction: Build models that use project plans to identify which of the tasks are the most critical to complete the project on time before the project even starts.
Project planning: Build models that accept blueprints and 3D Replit models to create project timelines and scaffolding as the starting point for human planners to make project planning a breeze.
Progress trackers: Build models that take in multi-model data from static cameras or videos from humans/drones on-site, voice notes, etc about a task to predict what percent of the task is complete.
Risk mitigation: Build, train and deploy models that can predict the best path forward for risk mitigation or timeline amelioration for major projects.
Continuous improvement: Evaluate models repeatedly and prototype methods and new models to improve the accuracy of current features, inspired by latest research publications.
What you'll do
Own features from technical design through release and beyond: measure adoption, learn and keep improving.
Work alongside Full Stack engineers to help them turn new models into production features customers rely on.
Partner with product, design and customers to find the problems most worth solving.
Tune performance across the modelling stack, from accuracy to latency and reliability.
Raise the bar on architecture, testing, observability and how the team ships.
Help shape the product and technical roadmap.
Requirements
You might be a fit if you
Have 3+ years of experience building and shipping Machine Learning models in production, with real ownership of the backend.
Have 1+ years of experience building and shipping VLM/LLM based features to production.
Are strong in Python on the backend (FastAPI, Django or similar).
Are solid at relational data modelling in Postgres: schema design, indexes, query plans, migrations, and an ORM such as SQLAlchemy.
Have shipped agentic workflows, RAG systems and high-volume data pipelines.
Have experience with async work: queues, background jobs, retries, idempotency, and what happens when a worker dies halfway.
Are hands-on with AWS (S3, RDS, a managed queue or event service, containers or serverless).
Write tests and debug production issues from logs, metrics and traces, not guesses.
Are clear in written communication and comfortable making pragmatic trade-offs in a small team that moves fast.
Bonus points
Experience with workflow orchestration for multi-step jobs (Temporal, Step Functions, Airflow or similar).
You've worked on a legacy-to-modern migration, or on products for regulated enterprise or public-sector customers (SOC 2, data residency).
Early-stage or founding-team experience.
None of these are required. If the core of this role excites you, apply.
Benefits
Why join now
Real stakes, real customers. Real stakes, real customers. You won't be building for a hypothetical buyer. Owners and contractors are building now, on deadlines measured in revenue per day.
Blank canvas, solid ground. You get first-engineer level scope on a product that already has paying customers, real data and proven outcomes to tell.
Defining a category.Agentic software for physical infrastructure is a new category. What you build will become the standard.
Small, senior team. Flat, direct, low on process. We debate openly, decide quickly and ship.
Compensation and location
Total Comp: $180,000-230,000, based on experience, level, and location.
Benefits: Health, Dental, Vision, 401(k), PTO .
Location: New York, San Francisco, and Montreal are preferred. Occasional travel to may be required.
