Applied AI Systems Engineer

About Disha We believe the future of healthcare is one where everyone carries a personal doctor in their phone—someone who knows them, understands their health journey, and is accessible whenever they need support. Disha is a step toward that future. Disha is an AI health coach that supports people through personalized, ongoing conversations across chat and voice. We are building it to make healthcare more proactive, continuous, and accessible—not limited to the few minutes someone spends with a doctor.Disha launched on September 20, 2025. Nine months later, as of June 20, 2026:-More than 500,000 people used Disha monthly.In May alone, users spoke with Disha for over 3.2 million minutes an exchanged more than 6 million chat messages.New users spent an average of 17 minutes talking to Disha on their first day.We recovered customer-acquisition and AI-inference costs within 90 days and are profitable.Disha is backed by Elevation Capital and General Catalyst. We are a small team of highly effective people—alumni of the IITs and AIIMS Delhi working alongside experienced healthcare professionals—who take ownership, move quickly, and care deeply about the quality of our work. The founders themselves bring 5+years of experience in care delivery.We are looking for an Applied AI Systems Engineer who will work closely with the founder to turn ambiguous product problems into reliable AI systems.We need someone who can sit between product thinking, agent architecture,experimentation, and implementation.What You Will Do:-Your work is one loop, run over and over until the system is reliable:Understand. Dig into real conversations, traces, and failures to pin down a problem in Disha's chat or voice experience.Design. Turn it into a system: agentic workflows, prompts, RAG, memory, tools, guardrails, structured outputs, or plain deterministic code. Decide what an LLM should handle and what code should enforce, and pick the model each piece needs by balancing speed and intelligence against cost.Prototype. Prove the design with quick prototypes and internal tooling — vibe coding encouraged.Evaluate. Build evaluation datasets and quality metrics, and let the numbers — not the demo — tell you whether it works.Hand off and own. Iterate until the design is solid, hand it to the tech team for production, and own how well it performs after launch.Around that loop, you will design how prompts, protocols, and guardrails get authored, reviewed, versioned, and deployed; document your decisions so others can build on them; and keep testing new models and techniques as the field moves.Who We Are Looking ForThe basics:1.Roughly 3-5 years in software engineering, ML, or data science. We care about the systems you have built and the decisions you made, not your degree or title.2.Strong Python and software-engineering fundamentals — enough to build prototypes, internal tools, and data pipelines quickly.3.Hands-on experience building with LLMs: prompt and context engineering,tool calling, RAG, memory, structured outputs.4.Experience creating evaluations and debugging live systems from logs, traces,and conversation data.