Supril Singh is an active participant in the AI agent ecosystem, specifically focusing on the developer tooling and infrastructure layer. His work is relevant because it implements modern agentic patterns like the Model Context Protocol (MCP) and LangGraph to solve complex software engineering tasks. By building tools like Traceminer and recursive codebase analyzers, he demonstrates how agents can be used for more than just text generation, moving into the realm of system analysis and automated reverse engineering.
In the broader agent stack, Singh operates at the intersection of agent orchestration and sandboxed execution. He champions the idea that agents should interact with software systems through programmatic inspection and code execution rather than just large-scale context pasting. This approach is critical for the development of autonomous software engineering agents that need to remain coherent across massive, multi-file codebases.
Supril Singh is a Pune-based engineer whose work centers on the premise that the next layer of software development involves building the machine that develops other machines. As a Senior Cloud Application Developer at ZS Associates, his professional focus lies in architecting production monitoring platforms and automated migration systems. However, his independent projects reveal a specific interest in the intersection of platform engineering and agentic workflows, particularly those that solve granular developer pain points.
His engineering philosophy is grounded in building production-ready systems rather than demo-only wrappers. This is visible in his choice of stack, which includes LangGraph for orchestration, MCP-style architectures for tool integration, and sandboxed TypeScript execution for security and reliability. These choices indicate a move away from the simple prompt-and-response pattern toward more complex, multi-step agentic behaviors that can operate autonomously within defined technical boundaries.
One of the more distinct projects in Singh's portfolio is Traceminer. The tool is designed to reverse engineer undocumented APIs by analyzing HAR files. Instead of requiring a human to manually map out traffic noise, Traceminer uses agentic workflows to reconstruct private API behavior and generate clean maps and SDKs. This application of agents demonstrates a practical use case for LLM reasoning: handling the messy, unstructured data of network traces and converting it into structured, testable code.
Singh also develops experimental tools for codebase analysis. His RLM-style (Recursive Language Model) Codebase Analyzer addresses the context window limitations of standard LLMs. By keeping the full repository outside the immediate model context and using programmatic inspection to pull in only the relevant files, the system can perform multi-file debugging and architecture reviews. The analyzer works in passes—locating files, summarizing behavior, and comparing call paths—to maintain coherence across large systems that would otherwise overwhelm a single reasoning loop.
Beyond developer tooling, Singh has built automation for creators, specifically a Telegram-first AI studio. This system manages the entire pipeline for short-form video production, from script generation and voice cloning to final renders. It uses a mix of local and serverless AI workers to handle GPU-heavy inference, managed through a queuing system that eliminates the need for manual oversight.
Singh's background includes a Bachelor of Engineering from RCOEM, Nagpur, and three years of experience at ZS Associates. His work there includes building "Smart Assist" chatbots for Snowflake and Power BI, along with license reconciliation systems that automate entitlement checks. This enterprise experience informs his independent work, leading to a focus on audit logs, RBAC (role-based access control), and structured outputs. He operates as an independent builder in the agent ecosystem, contributing tools that bridge the gap between high-level LLM capabilities and practical, day-to-day engineering tasks.
Reverse engineer undocumented APIs from HAR files using agentic workflows.