About Me

I build AI systems that work in the real world.

Applied AI Engineer based in Jaipur, India. Currently at Multilipi, leading AI product development and government AI projects. Previously at ISRO NRSC, where I worked on satellite data systems that handle terabytes of national geospatial data.

I started my B.Tech in Computer Science with an AI/ML specialization at JECRC University, Jaipur in 2021. I knew early on that I wanted to work on real systems, not just assignments and toy datasets. So I kept building - college projects on medical imaging, satellite land-use classification, and computer vision.

In 2023, I led my team to 3rd Rank at Smart India Hackathon (national level) with a micro-vibration amplification system for non-contact equipment damage detection using computer vision. That experience taught me how to build under pressure and how to communicate technical ideas to non-technical stakeholders - something that's served me enormously since.

In late 2024, I interned at Kapaxha Dynamics, a defence tech startup, where I designed software architecture for drone swarm control - custom UDP/TCP protocols for near-zero latency coordination, and YOLO-based leader drone tracking with Kalman filtering for smooth trajectory prediction. It was my first brush with hard real-time system requirements.

Early 2025 brought the opportunity I'd been working toward: a research internship at ISRO's National Remote Sensing Centre (NRSC), Hyderabad. I spent ~4.5 months working alongside ISRO scientists on geospatial AI problems at national scale - building data pipelines for petabyte-scale satellite imagery, designing a satellite data archival and monitoring system with Kafka-based event streaming, and developing POCs on DEM generation, multi-spectral classification, and intelligent image compression. We co-authored a research paper (pending formal submission).

Since July 2025, I've been at Multilipi, where I wear two hats simultaneously. On the core product side, I architected the company's first GEO (Generative Engine Optimization) system - enabling clients to be cited by ChatGPT, Gemini, Grok, and Perplexity. I also upgraded the translation pipeline to Gemini, cutting costs by 40%. On the government projects side, I lead a team of 3–4 developers delivering AI systems for NPA, Bhashini, and Bihar Government - including a live RAG-powered complaint processing platform, a website translation system, and an ongoing Kaithi script OCR project.

Hard problem, real impact - that combination is what I look for.

I'm drawn to problems where the technical difficulty is matched by the real-world importance. Kaithi OCR is hard because there's almost no prior work, no pre-trained model, and extremely limited data - but the documents being digitized are irreplaceable historical records. The NPA complaint platform is technically interesting because multilingual real-time audio processing is genuinely complex - but it's deployed at a national police training institution and actively used to train officers.

I'm also deeply interested in multilingual AI. Most AI systems are built English-first, and the gap for Indian languages is enormous. I've spent a lot of time working within the Bhashini ecosystem, using models that handle 32 Indian languages, and thinking about what it means to build AI that works for a country as linguistically diverse as India.

Education