Dhruv Varshney@dhruvbit
1
Events attended
1
Submissions made
India
1 year of experience
About me
CURRENTLY WORKING ON ALPACA HACKATHOON I’m a student and an independent developer interested in AI, software engineering, and building practical systems. I enjoy understanding how things work under the hood rather than just relying on existing frameworks. Currently, I’m working on AI agents and experimenting with local AI, automation, and intelligent tool-use systems. I’m especially interested in building things that are actually useful, scalable, and reliable — not just demos. Always learning, experimenting, breaking things, and building them better.
🤓 Latest Submissions
Axiom Trade Labs
Axiom Trade Labs: The New Standard for Autonomous, Risk-Adjusted Trading The integration of AI in financial markets often faces a critical flaw: giving Large Language Models (LLMs) direct execution authority is inherently dangerous due to hallucinations and unrestricted API access. Axiom Trade Labs solves this by introducing a Dual-Mode Hybrid Architecture that strictly separates AI reasoning from deterministic financial execution. With Axiom, AI proposes, but mathematics decides. Key Architectural Highlights: Deterministic Safety Pipeline: Every user command or autonomous proposal passes through our SCSV Router for semantic extraction, followed by an independent Risk Engine. This engine mathematically evaluates exposure limits, concentration, and daily loss halts. If safety criteria aren't met, the system intelligently defaults to a "NO_TRADE" state. Isolated CORE-X Execution: Execution is handled by CORE-X, a strictly isolated boundary ensuring safe, idempotent communication with the brokerage API. AI never directly touches the broker. Event-Driven Autonomy: Instead of pinging LLMs on every market tick (which is slow and expensive), Axiom uses high-speed deterministic filters to monitor live market data, waking the AI for deep reasoning only during material events. Bayesian Adaptive Learning: The system maintains a transparent Decision Ledger, tracking real-world outcomes to automatically adjust strategy confidence weights based on performance in current market regimes. Dual-Mode Interface: Users can let the background Autonomous Engine trade systematically or interact with the AI Trading Agent via a unified, real-time terminal UI. Axiom is engineered with strict "fail-closed" behavior. For example, runtime policy changes instantly trigger a 60-second "Uncertainty Freeze," pausing new executions to protect capital while state synchronizes. Axiom isn't just an AI trading bot; it is a secure, institutional-grade execution infrastructure built for the AI era.
4 Sep 2026
