Remote in Brazil+1 moreMore locations: São Paulo, State of São Paulo, Brazil
Remote
Software EngineeringBackend Engineering
About the job
Requirements
Have at least four years of experience as a backend engineer.
Have strong proficiency in Ruby on Rails and PostgreSQL; experience with another backend stack may be considered for an exceptional candidate.
Have experience designing data models and schemas that accommodate changing product requirements, and be able to discuss design trade-offs and lessons learned.
Have experience building or maintaining data-ingestion pipelines from unstructured external sources such as documents, third-party feeds, or vendor exports.
Reason rigorously with data, assess whether results provide enough evidence for customers to act, and recognize when evidence is insufficient.
Be comfortable building production systems with large language model or agentic APIs, including pipeline design, evaluation, and responsible handling of nondeterministic output.
Be willing to work autonomously and execute in an early-stage environment.
Have previous experience as a developer in high-growth startups.
Care about transparency in the products people buy and consume.
Responsibilities
Improve and rework the extraction pipeline that converts external laboratory reports into structured records using document artificial intelligence and large language model extraction.
Strengthen validation and correction processes so extraction errors are detected, surfaced, and corrected before data informs compliance decisions.
Build a taxonomy that makes results from different sources comparable, including unit normalization, cross-laboratory naming, method equivalence, detection limits, and source-specific reporting edge cases.
Enable cohort analysis across suppliers, products, and categories after data has been normalized.
Organize and query test results across time and sources to identify trends, shifts, out-of-specification risk, and variance by supplier or batch.
Ensure customer-facing conclusions are statistically defensible, including determining when there is sufficient signal for action and communicating when there is not.
Build agentic workflows over the data, designing what models can access, how outputs are verified, and how conclusions remain auditable.
Design data models that serve customers, internal operations, external partners, and compliance without maintaining separate versions of the underlying record.
Desired Qualifications
Experience with analytical data stores such as ClickHouse, BigQuery, Snowflake, DuckDB, or similar systems is a strong plus.
Familiarity with Turbo (Hotwire) and Tailwind is helpful but not required.
About Sierra Studio
When you apply to Sierra, you join an ecosystem. We specialize in connecting talented Brazilian professionals with exciting career opportunities in a highly-vetted small community of growing companies in the US. Applying to Sierra seamlessly integrates you into this ecosystem, granting automatic eligibility for all relevant opportunities we offer.
About our hiring partner
Our hiring partner is a venture-backed early-stage US startup building software and data infrastructure for product testing and compliance. Their customers are consumer brands and manufacturers who need to verify what is actually in the products they sell, prove it to regulators and marketplaces, and share it openly with their own customers.
Testing today is slow, expensive, and the results end up buried in PDFs that nobody can compare or act on. Our partner is building the layer that changes that: a platform where testing results become structured, comparable data — data that automates compliance work, and that brands can publish to differentiate themselves.
The team is small, technical, and founded by people who have built and scaled companies before. You'd be an early engineer with real ownership of a core part of the product.
Role Overview
This is a backend role with the weight on data. The product surface and the core pipelines already exist — this hire is about deepening them. The team's next set of bets are data problems: pulling in messy external data at volume, organizing it so it means the same thing across sources, and turning the resulting database into something customers can act on.
The stack is Rails end to end and there's an established design system, so front-end work here is light and mostly mechanical. You may pick some of it up, or it may be handed to someone else — either way, it's not the center of this role and it's not what you'll be evaluated on.
What you'll work on
1. External data ingestion
Customers arrive with thousands of historical test reports from external labs — all PDFs, every lab with its own format, none of it structured. There's already a working ingestion pipeline; a good part of this role is making it substantially better.
Improve and rework the extraction pipeline that turns those documents into structured records (document AI + LLM extraction)
Strengthen the validation and correction layer around it. Extraction that is 90% right isn't good enough when the output feeds compliance decisions — how errors get caught, surfaced, and corrected is as much of the job as the extraction itself
Build the taxonomy that makes results from different sources comparable at all: unit normalization, naming across labs, method equivalence, detection limits, and the edge cases in how each source reports the same measurement
Enable cohort analysis on top of that — comparing across suppliers, products, and categories once the data finally lines up
2. Turning the results database into something customers can act on
The platform holds a large and growing body of test results. Making that legible and useful to customers is the second half of the role.
Organize and query results across time and across sources: trends, shifts, out-of-spec risk, variance by supplier or batch
Make sure what reaches a customer is statistically defendable — the bar is knowing when there's enough signal for a customer to act on something, and being honest when there isn't
Build agentic workflows on top of that data. The interesting engineering here isn't calling a model — it's designing what it can touch, how outputs get verified, and how conclusions stay auditable
Data modeling across audiences
As the product's scope grows, the team keeps hitting the same wall: different audiences — customers, internal ops, external partners, compliance — look at the same underlying record and need a different shape of it. Designing a model that serves all of them without forking is a real, ongoing part of this job.
Stack
Ruby on Rails — the core application, and the center of their ecosystem
PostgreSQL — transactional data model
ClickHouse — one of their data warehouses, already in production for analytical workloads
Claude API — extraction and agentic workflows
Reducto — document parsing / PDF ingestion, already in use
Turbo (Hotwire) + Tailwind + an established design system — front end, when it comes up
Requirements
Experienced backend engineer (4+ YOE) with strong proficiency in Ruby on Rails and PostgreSQL. Rails is strongly preferred given the existing ecosystem, though a genuinely exceptional backend engineer from another stack is worth a conversation
Real experience with data modeling — you've designed schemas that had to survive changing product requirements, and you can walk through the trade-offs you made and what you'd do differently
Experience building or maintaining data ingestion pipelines from messy external sources: documents, third-party feeds, vendor exports, anything you don't control the format of
Analytical rigor. You don't need to be a data scientist or a statistician. The bar is that you reason well with data: you can look at a result and say whether there's enough there for a customer to act on, and you're comfortable saying "not yet" when there isn't. Think strong data analyst instinct, applied by an engineer
Comfort building on LLM / agentic APIs in production: pipeline design, evaluation, and handling non-deterministic output responsibly
Experience with analytical data stores (ClickHouse, BigQuery, Snowflake, DuckDB or similar) — strong plus
Familiarity with Turbo (Hotwire) and Tailwind — nice to have, not a filter
Bias to action — this is early stage, 0 to 1 execution
Excellent communication, comfortable working autonomously and seeing the big picture
Previous experience as a developer in high-growth startups
Mission alignment matters here: this team cares about transparency in what people buy and consume, and they hire for people who care about it too
Values
First Principles
Deal with ambiguity, deconstruct the problem, build the optimal solution
Question every requirement → delete any part or process you can → simplify and optimize → accelerate cycle time → automate
Standard of Excellence
High standards are contagious. A+ talent attracts A+ talent
There's a glut of mediocrity in the world. They're chasing the products and the people that strive for excellence
Bias for Action
Take initiative, decide quickly, experiment, and learn from failures
Customer expectations rise over time, which means improving every single day
Low Ego
Coachability. It's not about being right, it's about getting to the right answer
Reacts calmly to criticism, and treats feedback as information rather than a threat
Respectfully challenges decisions you disagree with, even when it's uncomfortable — and once a decision is made, commits fully and moves forward