A year as core engineer on autonomous agents that run inside SAP.

The problem

FlowGen Labs builds autonomous agents that execute long-running business processes inside SAP ECC and S/4HANA: order-to-cash, procure-to-pay, cash application, dispute resolution. The agents are grounded in the customer's enterprise ontology and operating procedures, run with human-in-the-loop controls on high-impact actions, and are priced on measured outcomes. FlowGen is an SAP Registered Partner; one public deployment reports more than $9M in annual process savings and went live in four weeks.

That kind of product has an unusual engineering surface. The agents have to hold state across workflows that run for hours or days, retrieve the right procedure from large document sets, act through SAP interfaces safely, and be deployable into locked-down enterprise environments. The founding team, from Stanford, Google DeepMind and Meta AI, needed senior engineers who could take a hard problem from design to production without hand-holding.

What we built

Techtribe co-founder Aashan joined as a core engineer for a year, roughly 900 hours. The work covered the full stack of an agentic enterprise product:

Agent and LLM systems. Building and hardening the agent loops that execute procedures: tool-calling contracts, structured outputs, retries and fallbacks, and the evaluation needed to trust an agent with a financial transaction.

Retrieval. RAG and vector search over enterprise procedures and documents, so agents ground their actions in the customer's actual operating rules rather than general knowledge.

Application layer. Python and FastAPI services on the back end, TypeScript and Next.js on the front end, for the surfaces operators use to configure, monitor and approve agent actions.

Infrastructure. AWS provisioned with Terraform, built for the deployment constraints of enterprise customers.

Developer tooling. A VS Code extension for the team's internal workflow.

The specifics of the architecture are FlowGen's to share; what we can say is that this is the exact shape of work Techtribe sells: production agents, retrieval, and the infrastructure to run them in a customer's environment.

Results

A year of continuous delivery as a core engineer on a venture-backed enterprise AI product, with a written recommendation from the CEO covering agents, RAG, infrastructure and full-stack ownership. Among all our engagements this is the one that most directly demonstrates the capability behind our positioning.