RedForge — AI Red-teaming & GenAI Eval Platform
Target-agnostic platform that attacks any LLM chatbot with 200+ OWASP-LLM-Top-10 test cases and an adversarial mutation engine. Built on personal time, adopted as a company product.
role: Creator & product lead
Story
Conceived, built and validated independently on personal time (May–Jun 2026). Pitched internally, secured company adoption in Jul 2026, and now lead it as a funded product with an expanding roadmap.
What it does
A target-agnostic harness: assess any LLM chatbot via direct API integration or browser automation — so even third-party assistants with no API can be black-box tested.
Coverage
- OWASP LLM Top 10 — direct & indirect/stored prompt injection, jailbreaks, system-prompt extraction, cross-customer data leakage, insecure output handling, excessive agency, sensitive-information disclosure, unbounded consumption.
- Mutation engine — base64, hex, ROT13/Caesar, homoglyphs, zero-width & Unicode-tag injection, leetspeak, roleplay, random augmentation; un-mutated baselines kept as controls.
- Responsible-AI & quality suites — groundedness, answer relevance, consistency, PII handling, fairness/bias, toxicity and refusal calibration.
Safe by design
Explicit authorization gating, per-scan request caps, target recon, and reproducible YAML scan configurations.
Tool-Selection LLM — LoRA Fine-tune on xLAM
Teaching a small LLM to pick the right tool/API for an agent — and say how confident it is. LoRA-tuned Qwen3.5-2B with Unsloth on xLAM function-calling data: 50.6% → 100% accuracy, ECE 0.220 → 0.002.
Agentic AI Chatbot Platform
Production multi-agent assistant for cybersecurity & case management — a router agent dispatching to five specialist agents across hybrid RAG, text-to-SQL and action execution, on self-hosted vLLM.
Self-learning Few-shot Text-to-SQL
A retrieval-driven few-shot engine that harvests successful queries as exemplars and injects them dynamically — lifting text-to-SQL accuracy from 60% to 95% with no retraining.