About the Role
This role is for a world-class systems engineer and builder to lead the development of a next-generation, multimodal AI assistant platform that enables secure, self-correcting agentic workflows. The candidate will act as a player-coach, balancing deep technical work with leadership of a small team of AI systems engineers.
Responsibilities
- Design and implement robust, multi-step agent state machines with self-correction, tool integration, and persistent long-context memory.
- Develop and maintain secure runtime environments for executing model-generated code, ensuring isolation and compliance with enterprise security standards.
- Lead the design of modular tool-calling systems using Model Context Protocol (MCP) and custom APIs to expand agent functionality.
- Create automated testing frameworks using LLM-based evaluation and trajectory analysis to validate agent behavior, tool selection, and error recovery.
- Optimize system performance through latency reduction, semantic caching, speculative execution, and real-time streaming of agent states.
- Recruit, mentor, and scale a small, high-performing team of software and AI engineers.
Tech Stack
Orchestration & Frameworks: LangGraph, LangChain, Go/Python, React, Node, Next.js, Infrastructure & Security: Secure sandboxed runtimes (Docker, gVisor, WebAssembly), GCP, AWS, Kubernetes, Cloud SQL (PostgreSQL, TimescaleDB), Redis, Pub/Sub, AI & Models: Integration with GPT-4o, o3, Claude 3.5 Sonnet, DeepSeek via function-calling, fine-tuning, and hybrid RAG pipelines, Standards: Model Context Protocol (MCP) tool schemas and APIs
Work Arrangement
Hybrid — San Francisco, Bengaluru
Team
Lean, fast-moving team focused on building advanced AI systems with minimal bureaucracy.
About Instabase and the Evolution of Our Business Model
The company enables large enterprises in banking, insurance, and healthcare to automate workflows involving unstructured data. Its platform is widely used for processing complex documents at scale. The organization is now advancing its platform with SuperApp, a multimodal, stateful AI assistant that supports autonomous, self-correcting execution. This new platform combines secure document understanding with real-time code execution, tool orchestration via MCP, and open-ended reasoning in a protected environment.
The Role: Player-Coach Leadership
This position is designed for a top-tier systems builder who can lead by example. The team structure is lean, emphasizing speed and technical excellence. The manager will function as a player-coach, spending 60% to 70% of their time on hands-on engineering tasks such as designing agent loops, writing production code, and building secure execution environments, and 30% to 40% on leadership duties including hiring, mentoring, code reviews, and aligning technical execution with product goals.
