Senior Founding Software Engineer, Ai Agent Systems

About AddBack
AddBack is an AI platform for pre-transaction financial due diligence. We ingest raw deal data (general ledgers, trial balances, ERP exports, bank statements, full data rooms) and produce institutional-grade deliverables: normalized EBITDA, EBITDA bridges, red flag summaries, IC memos, all cited and editable.
Our approach is reconciliation-first: balance sheet to general ledger to subledgers to bank. That's what separates real diligence from document summarization. The product is live, we're working on real deals with design partners, and we have a strong pipeline of enterprise prospects across private equity, M&A advisory, and private credit.
Our co-founders bring deep experience on both sides of the problem: Jason is a successful entrepreneur with extensive PE and deal experience, and our other co-founder comes from renowned tech and AI companies. You'll work directly with both.
Vision, Mission, Values
Vision. Due diligence hasn't changed in decades: slow, expensive, painful, and accepted as the price of doing deals. We're building the buyer intelligence layer that should have existed all along. Not to make diligence slightly better, but to change what buyers expect from it entirely.
Mission. Give every buyer the intelligence to walk into a deal clear-eyed in minutes, not weeks, at a fraction of the cost.
Values
Challenger. We took on something hard, and we challenge each other constantly and constructively. We don't want "sounds great," we want "here's where this breaks." The best idea wins, no matter who it comes from.
JEDI: Let's Just Effing Do It! Own it and move fast. A good plan fiercely executed now beats a perfect plan later. Ask for help early, unblock each other fast, and stay ruthless about what actually moves the vision forward.
Ask for forgiveness, not permission. Bias toward action. For everyday calls, assume you have the keys; for big or irreversible ones, do a quick gut-check. Own the outcome either way.
The role
You'll own our multi-agent pipeline end to end: classification, ingestion, extraction, reconciliation, and QoE analysis. The system works. Your job is to make it reliable, consistent, fast, and scalable to data rooms 10x the size of what we process today.
This is a founding-level role. You won't be handed tickets. You will join and lead our current engineering team. You'll decide what gets built, in what order, and why, working directly with the founders on product roadmap and technical strategy.
What you'll do
Own the agent architecture. Take over a production system of ~50 specialized agents and make its outputs consistent and reproducible run after run. Decide what should be an LLM call and what should be deterministic code.
Build the evaluation layer. Turn our proprietary dataset into an automated eval and regression harness, so every change is measured against ground truth.
Scale ingestion. Handle general ledgers of 13M+ tokens and messy multi-format data rooms with robust chunking, parsing, and deduplication.
Run the infrastructure. CI/CD, deployment, monitoring, alerting, and cost and latency optimization.
Fix what's broken, fast. Debug production issues on live client deals, find root causes, and make sure they don't happen twice.
Shape the roadmap. Turn customer feedback and GTM priorities into technical plans. Tell us what's realistic, what's risky, and what we should build instead.
Set the engineering bar. Code quality, documentation, architecture decisions, and eventually, hiring and leading the engineering team.
What we're looking for
We don't care about degrees. We care about what you've built, shipped, and owned.
Must have
5+ years as a software engineer, including 1 to 2 years building LLM-based systems in production (not just prototypes)
Hands-on experience with multi-agent or multi-step LLM workflows: orchestration, state management, retries, error handling, idempotency
Strong Python and backend fundamentals (APIs, async processing, job queues, Postgres)
A track record making LLM outputs reliable: structured outputs, schema validation, prompt versioning, evals, guardrails
Real DevOps experience: Docker, CI/CD, Azure, infrastructure as code, observability
Experience with large or messy data: document parsing, OCR, spreadsheet extraction, long-context and chunking strategies
Strong remote habits: you write clearly, share progress without being asked, and flag blockers early
At least 4 hours of daily overlap with Eastern Time, and fluent professional English
Nice to have
Orchestration frameworks
LLM observability and eval tooling
Tabular data tooling at scale
Fintech, accounting software, or financial reconciliation background
Security for sensitive client data: tenant isolation, encryption, SOC 2 readiness
Experience as one of the first engineers at a startup
Don't tick every box? Apply anyway.
If you don't have 5 years, or your title says "developer" rather than "engineer," but you've built real things and believe you can help us grow as fast as our pipeline, we want to hear from you. We promote fast: as AddBack hits its milestones, the people who got us there take on more scope, more ownership, and more leadership.
Why AddBack
Founding role with real ownership over architecture, roadmap, and the future engineering team
Significant equity and fast growth opportunities
Fully remote, with the option to work from our Montréal or New York offices
Live demand, real deals, and a proprietary dataset competitors can't easily replicate
Hard, interesting problems: accounting semantics, large-scale reconciliation, and making AI trustworthy enough to sign off on
Compensation will be competitive according to your experience + Benefits + Office Stipend + 31 days PTO.

Work arrangement
Yes