Title : AI Agent Engineer | Founding Team

Location: Bengaluru / Hybrid

Type: Full time | Founding Team

About DecisionX

DecisionX is building the Decision Infrastructure for the enterprise - a causal, self-learning AI system that helps organizations move beyond search and retrieval to actual decision-making. We're trusted by global enterprises across pharma, CPG, and manufacturing, powering 10,000+ decisions a month. Ranked #2 globally in enterprise reasoning on the Spider 2.0 Lite benchmark and recognized by NVIDIA as an AI startup to watch - we're a small, fast-moving team solving one of the hardest unsolved problems in enterprise AI.

Why this role exists ?

We are live inside CPG & Pharma field-force workflows, and manufacturing RCA chains, and every one of those go-lives needs an engineer at heart who can sit inside the customer's world.

What you'll do ?

- Take a customer's raw mess: ERP exports, flat files, half-documented spreadsheets, and turn it into a working worldview of their business.
- Join operational datasets across systems that were never meant to talk to each other
- Write and run the eval query bank for each customer's real questions, on their real data, and fix the agent when it's confidently wrong.
- Estimate LLM cost and latency for a customer's actual query volume before it becomes a support ticket.
- Sit in requirement calls with customers, and come out with a spec they sign off on.
- Own the pilot: timeline, data quality,



agent accuracy, go-live.
- Notice when three customers hit the same wall, and push the fix into the platform once.

What you'll bring

- 1-3 years as a software or AI/ML engineer. Founding engineer or early-startup experience counts for more than the years.
- SQL fluency across joined, messy, multi-table enterprise data. You've done this on data that wasn't clean and wasn't yours.
- Strong Python, and enough JS/TS to not be blocked on a frontend engineer.
- You've built with LLMs in a real product: prompting, RAG, tool-calling, agent orchestration. Not just a weekend hackathon, a thing that ran in front of a user.
- You've written evals before, and you know the difference between an agent that sounds right and one that is right.
- You can hold a technical conversation with a customer's IT team without needing a product manager translating [read: babysitting] for you.

Even better

- You've priced or optimised LLM cost and latency for a live product.
- You've familiarity with Databricks, Snowflake, or SAP systems.
- You've been the second or third engineer somewhere, or founded something yourself or worked on passion projects on the sides.

If your idea of a good week is a clean sprint with well-scoped out tickets, this role will annoy you. If it's solving a customer's problem by Thursday because a real business decision is riding on it, apply.

📌 AI Agent Engineer (Bengaluru)
🏢 DecisionX AI
📍 Bengaluru