The Engine Room

Built for Deterministic Execution

We build translation infrastructure converting raw corporate events into clean options and equity signals without AI hallucination.

Our Core Philosophy

Most modern fintech portals feed raw filings into general-purpose LLMs and execute trades based on black-box sentiment. In live markets, this approach is capital suicide. Optimistic press releases and complex CEO supply-chain commentaries trigger false positives, leading to rapid drawdowns.

No Black Boxes

At Candlery, we treat language models strictly as parsing compilers—never as decision makers. The AI's only job is to translate unstructured text into standard, structured JSON intent payloads.

The Hard-Coded Firewall

Every trade signal must pass through our hard-coded, non-AI Quantitative Risk Firewall. If a parsed intent lacks historical precedence or violates safety gates, it is silently rejected. We prefer missed opportunities over capital destruction.

System Comparison
Step 1: Raw InputFiling Scraped & StreamedFiling is parsed by custom crawlers. Real-time scraping intercepts news within milliseconds.
Step 2: Parsing GateNLP Intent CompilerLLM parses text strictly into structured parameters: { capex_increase: 0.12, risk_disruption: false }.
Step 3: Verification GateHard-Coded Quantitative FirewallIntent parameters checked against 5-year historical backtest analogs. Rejects any non-precedented signals.
Step 4: FIX Signal OutputZero-Hallucination Execution compilerClean execution signals routed directly to API gateways via FIX/JSON with custom stop-loss parameters.Verdict: Secure Trade Routing

Our Core Principles

Deterministic Firewalls

We refuse to trade on black-box sentiment. Every extracted signal passes through code-level verification gates, matching with mathematical historical precedents before order compilation.

Sub-Millisecond Ingestion

Our ingestion crawlers bypass standard browser pipelines, fetching BSE/NSE filings and SEC disclosures at the wire to parse corporate disclosures before they impact market quotes.

Developer-First Architecture

Built by systems architects and quantitative engineers, the Candlery platform exposes high-fidelity sandbox APIs, detailed JSON event structures, and direct FIX connectivity.

Our Journey & Iterations

Event-Driven Signal Engine (V3)

The current state of Candlery. We expanded the deterministic parser to support Futures & Options (F&O) hedging strategies, compiling corporate news directly into complex options legs.

We added real-time scraping nodes for India-specific filings (BSE/NSE announcements), launched the institutional cockpit interface for auditor reviews, and proved the Event-Drift strategy backtest at a 2.42 Sharpe Ratio.

V3 Systems Log (2026)
Telemetry Latency:12ms (Wire-to-Parse)
Supported Assets:Equity, F&O Derivatives
Active Win-Rate:64.2%
Event Backtest Sharpe:2.42 Sharpe

People

Charan Dubey
Founder, CEO

Charan founded Candlery to establish a zero-hallucination execution standard for retail and institutional traders in India. With a background in systematic software engineering and quantitative data parsing, he steers the platform's vision of converting corporate announcements into machine-ready signals. Outside of Candlery, he is an avid reader and table tennis enthusiast.

Systems DesignQuant ResearchCorporate DataStrategy
Pavani Janjanam
Co-founder and CFO
Financial PlanningCapital AllocationRisk OperationsCompliance
Archit Somayajula
CTO
Execution CompilerGoFIX ProtocolRisk Gates