Year
2025 — Present
Field
Space systems · Deep learning
Projects built & benchmarked
Degree programmes in computer science & data science
Real conjunction events studied
Projects
Selected work.
View all projects (3)
The stack
behind the work.
From choosing the architecture to shipping the interface. The tools I work with, and the foundations underneath them.
01Languages
The syntax behind the systems.
02AI & deep learning
From a first baseline to a model worth trusting.
03GenAI & agents
Context, retrieval and systems that can take the next step.
04Data & analytics
Turning messy inputs into useful evidence.
05Web engineering
Interfaces and services around the intelligence.
06Databases & tools
The infrastructure that keeps the work moving.
Beyond
the commit.
Building a system takes more than a model. It takes decisions, direction, and people moving together.
Dec 2024 — Apr 2025
Co-Founder
Niti Capitals · Wadhwani Foundation Ignite Cohort
From an idea to a financial product roadmap.
- Designed a retail wealth-management platform combining alternative assets with automated debt servicing.
- Formulated unit economics, multi-tier risk allocation models and product roadmaps.
- Completed the Wadhwani Foundation Ignite cohort with certification.
Sep 2024 — Jan 2025
Super Coordinator · Cultural Department
Paradox Fest · IIT Madras
Building the team behind the experience.
- Managed a 55-member team.
- Increased online registrations by 32.5% and achieved 96% turnout.
- Managed a budget of ₹80,000+ and curated 10 final events from 200+ proposals.
Certs, Courses
& Curiosity
A few things I’ve learned.
Agentic AI Certification
OracleCompleted
Agentic AI
AWS Academy Graduate · Cloud Architecting
Hedera Hashgraph Certification
The Hashgraph AssociationCompleted
Distributed ledger technology
Still connecting the dots.
Generative AI Certification
OracleIn progress
Expected October 2026
Questions,
experiments &
paper trails.
A little less jargon. A little more “so what?”
Research and publications, in plain English.
OrbitalGuard: learning to spot a close call
Can a model help distinguish a genuinely risky spacecraft encounter from another false alarm?
Spacecraft share an increasingly busy neighbourhood. OrbitalGuard studies close approaches between objects and compares several deep learning approaches for estimating collision risk.
The work benchmarks recurrent, convolutional and Siamese architectures on real conjunction events, with attention to missed high-risk encounters, false alerts and stable training. A companion 3D simulator makes the orbital scenarios easier to explore.
The project includes a 35-page IEEE-format technical report describing the experiments and their trade-offs. Publication is in progress; this entry is a research note, not a published or peer-reviewed paper.
Deep learning / Space systems / Collision-risk modelling