Sandeep Singh

At AiRA, I am building the brain of a personal AI agent. I don't just connect APIs; I solve the hard engineering problems required to make agents reliable. I architected the flagship "Daily Brief" system, a feature that digests emails, calendar events, and user preferences to deliver proactive morning updates. My focus is on production reliability and security. I solved complex distributed system challenges (like global timezone orchestration) to ensure tasks execute precisely when intended. I also worked on the implementation of CASA security standards, building token denylists in Redis and sanitizing injection vectors to make the system enterprise-ready. On the integration front, I built a deep bi-directional sync with WhatsApp, allowing the agent to understand reply contexts, monitor groups, and interact naturally - making AiRA not just a chatbot, but a proactive assistant.

At Indigle, I learned how to scale. As Co-founder & Founding Software Engineer, I architected and shipped Samagra (IndieRP), a full-stack campus ERP system that grew to 10,000+ active users. This wasn't a toy project. I built the entire infrastructure from scratch using React Native for mobile and FastAPI/Node.js for the backend. The biggest challenge was cost; I engineered a zero-cost architecture that handles 150k+ monthly API requests with sub-100ms latency. I also built real-time synchronization features (using Socket.io) that keep thousands of students updated instantly. This experience taught me how to handle production databases (MongoDB/Postgres), manage 99.9% uptime, and pivot based on real user feedback.

At IIT Kharagpur, I didn't just use LLMs; I optimized them. As a Research Intern, I tackled the problem of Multimodal Reasoning for IoT. Standard models like ChatGPT fail completely when asked to generate precise Arduino circuit wiring. To solve this, I built CircuitSynth. I fine-tuned open-source models (Qwen 4B/8B) using LoRA and QLoRA, creating a specialized pipeline that could "think" in hardware schematics. The result? A system that achieved 75% accuracy on tasks where base models scored 0%. This experience taught me the deep internals of Transformers, the art of efficient fine-tuning (PEFT), and how to deploy heavy models on constrained edge hardware.

• Conducted an intensive 12-hour workshop on Android development and React Native, attended by 100+ students from diverse technical backgrounds. • Maintained a high engagement rate with over 80% attendance retention until the end, reflecting strong delivery and content value. • Designed a hands-on curriculum covering mobile app fundamentals, UI/UX best practices, and deployment strategies. • Provided live coding demonstrations and one-on-one mentorship throughout the session, addressing real-time queries and project guidance. • Organized and managed a college-wide Hackathon involving 120+ teams, overseeing event logistics, technical support, judging criteria, and sponsor coordination. • Ensured a smooth and inclusive experience for all participants, from registration to final presentations, fostering innovation and collaboration. • Received positive feedback from both attendees and faculty, recognizing the initiative for its

• Architected the complete backend and frontend infrastructure across multiple modules, ensuring 100% uptime and high scalability. • Built and integrated secure authentication from scratch, including support for token-based flows. • Implemented robust rate limiting to defend against DDoS attacks and maintain service reliability. • Developed a granular authorization system for admins and students to ensure secure access control. • Created and launched the Samagra App, now serving over 5,000 monthly active users. • Optimized media storage by implementing object-based solutions using AWS S3, Cloudinary, and Firebase—significantly reducing storage costs. • Designed the system architecture for the entire codebase, focusing on performance, modularity, and maintainability. • Led a team of junior developers in building the frontend for the MMMUT Reso website. • Integrated Python-based scrapers to fetch real-time attendance data directly from the college's internal services. • Achieved a 5/5 user satisfaction rating from 99% of users, reflecting high user trust and engagement.

Spearheaded the development of 6 commercial web applications using the MERN stack, cutting client dev costs by 30% and reducing time-to-market by 45%. Built high-performance landing pages with HTML5, Tailwind CSS, and ES6+, boosting Google PageSpeed scores (95+) and increasing user engagement by 40%. Designed scalable REST APIs and MongoDB schemas supporting 50,000+ monthly requests with 99.9% uptime and <100ms response times. Integrated CI/CD pipelines and testing workflows, slashing deployment time by 60% while maintaining 98% code coverage.