Bitla Sravya Reddy
AI Engineer | AI Red Teamer | SIH’25 Finalist | TOP 10 ISRO Antariksh Bharatiya Hackathon ‘25
Hyderabad
bitlashravyareddy@gmail.com
I build AI. I break AI. I study what breaks. I’m Sravya Reddy, an AI/ML Engineer working at the intersection of AI Engineering, Agentic Systems, and AI Security. Currently, I work on red-teaming banking and healthcare AI agents using PyRIT, Garak, DeepTeam, and Promptfoo to uncover jailbreaks, prompt injection, data leakage, unsafe behavior, and tool abuse. I also work on agent governance, MCP security, policy-based tool permissions, and automated red-team workflows. From building FastAPI backends and browser-based agents to designing ML systems and automated red-team pipelines, my work has gradually moved toward one question: Can this AI system actually survive the real world? Because testing an LLM is one thing. Testing an agent that can think, call tools, access data, and take actions is a completely different battlefield. I’m interested in building AI systems that are capable, secure, explainable, and resilient by design.
AI Engineer
Truviq Systems
- Engineered a durable AI red teaming service using Temporal, automating adversarial testing aligned with NIST, OWASP, and ISO safety standards.
- Red-teamed banking and healthcare agents using PyRIT, Garak, and Promptfoo to mitigate prompt injection, jailbreaks, and data leakage risks.
- Built a centralized Agent Registry pipeline to securely validate, register, and manage the lifecycle of internal AI agents and MCP tools.
- Implemented policy-based governance via Dogwood-Cedar, enabling dynamic control and restriction over agent tool-calling permissions.
Machine Learning Engineer Intern
The Baronix
- Built a multi-head fashion classifier on a 250K-image dataset, hitting 89% accuracy and a validation loss under 0.20.
- Red-teamed the model via IBM ART using PGD and HopSkipJump attacks to uncover vulnerabilities to adversarial perturbations.
- Boosted model robustness by integrating adversarial insights and refined preprocessing into the training pipeline to mitigate attacks
AI Backend Developer Intern
Prodigal AI
- Engineered performance optimizations for a social media automation FastAPI backend, reducing p95 latency from 890ms to 230ms for top tenants and streamlining scheduled content delivery for marketing team.
- Applied STRIDE threat modeling across components to secure data flows and patch system vulnerabilities.
- Created a GEO Score Analyzer to audit internal brands against competitors across multiple sources, helping marketing leads identify blind spots in brand visibility and online presence.
- Implemented a browser-based agentic system that navigates multiple websites, extracts relevant information, and delivers structured data to the data team for downstream analysis
Red Team Leaders
Aug 2026Secure AI: Red Teaming & Safety Filters
Coursera
Aug 2026Neo4j GraphAcademy
Jul 2026Non-Human Identity Fundamentals Certified Professional
Oasis
Jul 2026
Attendee · Oct 24, 2026