Website:
Job details:
Most enterprise AI never reaches production. Gartner reports that 72% of CIOs are breaking even or losing money on their AI investments. We started Pragyaa to change that.
We build the platform that runs a company's back office on AI: Legal OS, Finance OS, compliance agents, and an ROI dashboard that shows exactly what the AI delivered. Under all of it sits one ontology, a shared model of the business that connects contracts, ledgers, entities and obligations.
This is the kind of result it produces. One of our first agents started at 30% accuracy using a frontier model straight out of the box. We added a knowledge graph and got to 70%. Evals, tuning and a live ERP integration took it to 90%. Today it has audited 600+ certificates on its own and gives a 75-year-old real estate company 20+ hours back every week. We now want to do that for every workflow, in every function, and we need the person who designs the system that makes it repeatable.
What you'll do
You'll architect the platform, write the pieces that matter most, and take work from design through to production:
- Design the core platform: the ontology, connectors, agent runtime, model gateway and evaluation layer shared by Legal, Finance, Compliance and ROI reporting.
- Write the critical code yourself: the ontology layer, extraction pipelines, orchestration and gateway.
- Review and integrate the team's work so every agent works end to end: typed, permissioned, auditable, with human approval where it matters.
- Run evaluation-first engineering: golden datasets from real cases, then continuous improvement on accuracy, cost and latency with every model release.
- Spend about 30% of your time with customers (general counsels, CFOs, compliance heads) turning their hardest workflows into architecture. The other 70% goes into our own platform, so each engagement makes the product stronger.
You might be a fit if
- You've spent 8+ years building software, including 4+ years building AI systems that real enterprise users depend on.
- You've designed a knowledge graph or ontology, and you know when a graph beats vector search and when it's overkill.
- You've shipped LLM agents or document-extraction systems, and you know how to get from 30% to 90% accuracy.
- You think in cost, latency and accuracy together, and you can cut one without breaking the others.
- You're comfortable with cloud platforms, APIs, MCP and messy enterprise systems (ERP, CRM, Microsoft 365, Slack).
- You'd rather ship than write slide decks, and you can explain your trade-offs to a CFO.
Why join us
- You own it: you'll be the architect of the platform, working directly with founders who built AI at Google, AWS, Azure and Meta.
- It's live: our agents already run legal, finance and compliance work for enterprise customers, from a crypto exchange to a 53-year-old retailer. You'll see your work used in production within weeks, not years.
- Results you can measure: every agent you ship reports hours saved and dollar ROI against the customer's own baseline.
- Frontier tools: we're a member of the Claude Partner Network and use AI agents across our own engineering, from spec to code to release.
- Momentum: we're three years in, with $3M+ ARR and 50+ people across the USA, Canada, the EU and APAC, and we're growing fast.
Click on Apply to know more.
