I am Prasun Kundu, an AI-native Full Stack Engineer and founder-minded developer. I architect autonomous agent workflows, build high-performance data backends, and scale product MVPs with end-to-end engineering ownership.
A real-time snapshot of platforms and architectures currently occupying my main terminal sessions.
Engineering a high-throughput prompt tracing and latency diagnostics dashboard for multi-agent chains.
Building autonomous web scrapers that digest prospects' company sites to output deeply-personalized B2B mails.
Investigating isolated sandboxes where autonomous agent swarms self-correct code to satisfy repo-wide test errors.
I operate with complete ownership across the lifecycle of software systems. Rather than viewing engineering as a task-completion role, I approach it from a product-first perspective: how does this code improve operations, unlock revenue, or enhance user agency?
My technical focus centers on the convergence of AI integration and systems engineering. I design autonomous multi-agent pipelines (hierarchical planning, RAG, and self-correcting workflows) and couple them with robust backends (Go/Node.js, high-speed clickhouses, caches) and modern web clients.
Having shipped complex client projects and developer tools, I understand what it takes to iterate at startup velocity without introducing instability. I implement comprehensive automated testing systems (API routing and visual E2E playwright rigs) to ensure our code remains solid across every deploy.
How I write code, design products, and architect production systems.
Technology is purely an instrument to solve user and business pain points. I align engineering objectives directly with product goals, ensuring every API, schema design, and UI flow adds immediate, measurable utility.
Shipping MVPs quickly is critical to get real-world user feedback. However, speed shouldn't invite regression errors. I build comprehensive test automation layers (Playwright and unit mocks) to shield codebases during rapid iterations.
Code isn't complete until it is fully observable in production. Tracing inputs/outputs, counting tokens, tracking nested chain latencies, and logging rate limits are vital components that must be architected from day one.
I prioritize modular, clean, and well-documented TypeScript and Go configurations over highly hyped libraries. I optimize for simple data layers, structured JSON routing, and containerization so systems scale cleanly.
Grouped by engineering layer. No progress bars, just production-grade competence.
Crafting premium user journeys, lightning-fast rendering states, and accessible layouts.
Architecting secure, modular APIs, WebSockets nodes, and high-performance routing patterns.
Designing agentic workflows, structured parsing, prompt tracing, and vector indexing.
Designing vector embeddings databases, telemetry archives, and caching structures.
Configuring workflows, sandboxed runs, and micro-deployment processes.
Ensuring 100% build reliability with E2E automation rigs and regression sweeps.
An AI-powered landing page auditor that uses vision-based models to analyze conversion friction points and copy flaws instantly.
An industrial assistant integrating live telemetry, thermal maps, and diagrams to diagnose overheating hardware issues.
Automated outreach personalization. Crawls prospects' domains to draft deeply contextualized outbound sequences.
A custom multi-agent orchestration framework supporting planner-executor patterns, memory routing, and tool correction loops.
An observability and debugging hub designed specifically for LLM applications. Traces multi-agent prompts, tracks intermediate API latency, and logs cumulative tokens.
Current copilot systems operate on short, single-file horizons. AgentStorm explores a future where autonomous agent swarms (planners, coders, testers, and reviewers) work in unison on containerized codebases.
Tailored engineering engagements focusing on rapid launch, robust backends, and agent integrations.
Integrating advanced LLM logic, structured JSON outputs, prompt chains, and custom RAG retrieval configurations.
Taking products from Figma mockup to live deployment inside a few weeks with robust user authentication and auth rules.
Engineering robust React/Next.js client workspaces tied with Node.js servers and highly-scalable database grids.
Designing administrative monitors, pipeline log analytics dashboards, and search consoles.
Structuring modular REST/WebSocket routers featuring rate-limit limits, robust session tokens, and caching.
Constructing background web crawling sequences and scheduled tasks pipelines that strip out manual routines.
Drafting systematic visual Playwright scripts, API load assessments, and automated regression setups for deploys.
Conducting codebase audits, scoping system architectures, mapping out database scaling, and LLM evaluations.
SDK and ClickHouse ingestion scripts for tracing prompt workflows, tokens, and latency limits.
Isolated Docker sandbox execution test runner exploring automated git commits self-debugging cycles.
3+ Years of end-to-end full-stack software execution and shipping products.
6 Shipped applications mapping AI analyzers, industrial WebSocket bots, and developer tools.
Next.js, React, Node.js, Go, Python, LangChain, Supabase, Playwright, ClickHouse, Docker.
Active systems code contributor on Radar SDK, AgentStorm Core, and Mega AI.
AI-Native Full Stack Engineer & Product Builder
Operating as an independent contractor, taking complete product-level ownership of SaaS ideas for early-stage startup clients. Architecting and deploying:
A vertical trace mapping core architectural growth, tools shipping, and current active research.
Initiated developer path building custom CLI modules and structuring custom JSON-RPC backends, learning HTTP networking, and setting up Linux deployments.
Combined multimodal Gemini/Claude APIs with Next.js dashboards. Deployed live models analyzing conversion metrics and high-throughput WebSocket streams for industrial telemetry monitoring.
Began operating as an independent contractor, taking full-stack ownership of SaaS ideas for early startup clients. Managed layouts, auth setups, API specifications, and visual test suites.
Faced prompt tracing loops in complex orchestrations, initiating developer work on Radar. Engineering Go micro-services and Redis Pub/Sub pipes to stream prompt analytics to ClickHouse.
Exploring multi-agent structures executing automated git tasks inside sandboxed containers, analyzing recursive testing logs and error corrections.
Whether you are a startup founder looking to launch a SaaS MVP, an engineering team looking to integrate agentic AI pipelines, or a CTO looking for a robust full-stack contractor—I am ready to take complete ownership.