Akasha A Prasad

Akasha leads Aevrin as CEO and CTO: the product architecture, the business strategy, the frontend and the core AI red-teaming pipeline, all built from the ground up. Aevrin audits AI chatbots for prompt injection, jailbreaks, sensitive-data leakage, policy failures and harmful outputs, so businesses can ship assistants that stay safe and compliant.

He started early. At fourteen he was on stage at the Indian Institute of Science presenting Metasploit-based penetration-testing techniques. Since then he has learnt by shipping: more than thirty real-world projects with over two hundred live users, five hackathon wins across robotics, AI security and venture sprints, and a first-year B.S. in Data Science at IIT Madras alongside Computer Science at Scaler School of Technology.

Built

• Aevrin: an AI security platform that red-teams chatbots before their users do, from the architecture to the audit pipeline.
• Mnemosyne: operational memory for autonomous SRE agents. A causal-graph engine that hands an agent root cause, similar past incidents and proven fixes in under two seconds, matched by behavioural fingerprint rather than service name, so it survives renames and topology drift with zero external infrastructure.
• LexGuard: a five-agent Gemini 2.5 pipeline that extracts contract clauses against the CUAD legal taxonomy and runs two-pass adversarial risk scoring. Live on GCP Cloud Run. Third place at Google PromptWars 2024.
• ClinicalTrialAgent: the first reinforcement-learning environment that teaches LLMs to design FDA/ICH-compliant clinical-trial protocols, scored by a composable reward across completeness, regulatory validity and scientific coherence. Selected for Meta’s hackathon.
• GradeTrack: a full-stack PWA for grades, assignments and progress across the IIT Madras BS programme, live with 180+ active students.
• Vector + graph native retrieval: a hybrid engine pairing FAISS/Chroma vector search with graph traversal for richer, faster answers on multi-hop questions than a vector database alone.
• Deepfake Engine Detector: a CNN forensic pipeline that names the generative engine behind synthetic media. Second place at a competitive hackathon.

Recognition

• Third place, Google PromptWars 2024 (LexGuard)
• Third place, VC Sprint Hackathon, organised by Campus Fund
• Fourth place, Robotics Hackathon: quadruped locomotion on the Unitree Go2
• Second place, AI Hackathon: deepfake engine detection
• Second place, Launchpad Hackathon, for a business pitch
• Top 14, Compassathon, hosted by IIT Madras, pitched on site
• Speaker at IISc on Metasploit-based penetration testing

Experience

Automation Intern at Decision Pinnacle (2026): built automated lead-generation pipelines for client businesses, combining API integrations, Python back ends and web crawlers, and automated the end-to-end sourcing and qualification workflow. Organiser, logistics and management, for Manch of Traders, the flagship event of Paradox, the IIT Madras fest.

Stack

React, TypeScript, Python, FastAPI, Java and Spring Boot, ML and RAG pipelines, FAISS and Chroma, SQL and MongoDB, Docker, GCP Cloud Run, Supabase, Gemini. Ethical hacking and cybersecurity throughout.