Dec 2025 - Present

AiRA

AI Engineer

• Standardized and optimized the Task Executor Engine for global users by implementing robust timezone logic and UTC normalization, achieving 100% accurate task scheduling and execution across all timezones.
• Developed a project reset automation script, reducing development and testing cycles by 20%.
• Engineered the flagship "Daily Brief" feature, automating personalized morning summaries of events, emails, and interests—now a core product highlight.
• Designed and shipped a user style adaptation pipeline, enabling the LLM to mimic user communication styles and context, driving high user engagement.
• Built advanced WhatsApp integrations: enabled reply context awareness, direct message deletion, group rule automation, and enhanced group moderation via two custom agents and memory files.
• Integrated ExaAI and Perplexity for context-aware, sanitized web search, including unicode sanitization for robust agent responses.
• Enforced CASA security standards: normalized MongoDB collection names, implemented Redis token denylist for secure logout, and sanitized email headers to prevent injection attacks.
• Optimized backend performance: converted query keys and group chat IDs to ObjectId for O(1) retrieval, dramatically improving message and group handling speed.
• Architected and launched the entire frontend for group rule automation, monitoring, account deletion, and secure authentication routes.
• Reduced pre-LLM token cost by 82% and improved overall system efficiency through prompt engineering and pipeline optimizations.
• Created group memory features and tools for agents to query and filter memory, enabling advanced group and rule moderation.
• Delivered measurable impact: increased system reliability, security, and user satisfaction, with flagship features widely adopted and praised by users and stakeholders.

May 2025 - August 2025

IIT Kharagpur

AI Research Intern (Multimodal AI)

• Developed "CircuitSynth," a multimodal reasoning framework fine-tuning Qwen models (4B, 8B, 1.5B) using LoRA/QLoRA to generate Arduino circuit JSONs.
• Achieved 75% accuracy on complex IoT circuit generation tasks, compared to 0% for base models like ChatGPT and Gemini, using a multi-stage (Coder -> Compressor -> Generator) pipeline.
• Implemented RAG retrievals and structured JSON validation to ensure executability of generated microcontroller code.

Oct 2025 - Present

n8n

Open Source Contributor

• Fixed critical issues with Webhook sandbox CSP properties (allow-same-origin missing), resolving form redirection failures.
• Diagnosed and fixed Windows build path failures for pnpm, improving the development experience for Windows contributors.
• Added comprehensive test cases to the core codebase to validate fixes and prevent future regressions.

Dec 2024 - Present

Indigle

Founding Software Engineer

• Architected and launched 'IndieRP,' a comprehensive full-stack ERP system that increased administrative efficiency by 40% across student, teacher, and administrator workflows
• Engineered real-time synchronization features utilizing Socket.io and Firebase Authentication, resulting in 99.9% uptime and securing over 10,000 daily user sessions
• Developed and shipped a responsive front-end using Next.js, implementing 25+ reusable components with Tailwind CSS and Shadcn/UI.
• Designed and optimized MongoDB database architecture, reducing query response time by 65% while managing 1TB+ of academic data.

Aug 2023 - Aug 2024

Web Accuracy

Software Developer Engineer

• Full-Stack Development: Designed, developed, and deployed full-stack websites and landing pages for six companies, achieving a 30% reduction in development costs.
• Frontend Optimization: Built responsive landing pages using HTML, Tailwind CSS, and JavaScript, significantly improving user engagement metrics.
• Backend Development: Developed scalable backend systems using Node.js, Express.js, and MongoDB.