.png&w=256&q=75)
1
1
India
1 year of experience
AI & Machine Learning student at Lovely Professional University with a passion for turning complex ideas into real products. I enjoy building autonomous AI systems, distributed applications, and high-performance software. Currently developing OmniCore, an AI-powered escrow platform for modern commerce, while exploring explainable AI and enterprise-scale systems. Open to collaborating with builders who enjoy solving meaningful problems.

Cross-border trade runs on a trust problem: the seller says pay first, the buyer says ship first, and the escrow meant to fix it is slow and manual. OmniCore automates exactly the verification work — and only that. The AI never has authority over money: it answers one question — can this transaction safely proceed automatically? — and routes every escrow to RELEASE (auto-settles through a C++ engine at ≥90% confidence), PENDING (missing evidence is requested), or DISPUTE (a case file is prepared for a human arbitrator, including the parties' immutable negotiation transcript as evidence). A deal starts with mutual assent: the buyer drafts, both sides negotiate on-platform, the buyer revises, and when the seller accepts, the contract is hash-frozen and funds lock at that moment. Before locking, a compliance agent flags corridor gaps (China→Zimbabwe in the demo). Uploaded PDFs pass a deterministic forensic layer — editor footprints, timestamp mismatches, save trails, math sanity. Four agents run live on gpt-oss-120b, served on AMD GPUs via Fireworks AI exclusively (a built-in mock mode runs offline without a key). In testing, the live verification agent caught a date-sequence flaw in my own demo data. Everything demonstrated works — built solo in six days. Roadmap: private deal rooms, dispute windows before release, a verified corridor-rules knowledge base, M-of-N release policies, and a guided contract builder with pro-rata remediation math.
13 Jul 2026