Pradyoth Prashanth

JPMorganChase

AI/ML Data Platforms Product Security Lead · Cybersecurity & Tech Controls

Feb 2026 — Present

Bengaluru, India

Security Engineer II

Top contributor on JPMC's Gen AI red team — running adversarial testing across 8+ enterprise AI engagements and designing the manual methodology the team uses against agentic systems.

• Run adversarial testing across 8+ enterprise AI engagements using PyRIT, Promptfoo, garak, and vendor scanning tools — 10,000+ automated prompts per use case (malicious plugins, encoding jailbreaks, black-box and gray-box agentic tests) plus 2,000+ manual prompts across multiple turns per target.
• Led assessments for Global Private Bank, Investment Bank Payments, Digital Banking, and the AI/ML Data Platform — surfacing issues severe enough that systems were re-architected before launch, and over 75% of all team findings, spanning prompt injection, encoded prompt bypass, system prompt leakage, guardrail architecture disclosure, and RAG overexposure.
• Consistently the top contributor on red team assessments — uncovering more findings than the rest of the team combined.
• Designed the manual assessment methodology the team uses against agentic architectures — targeting privilege misuse, intent misalignment, and multi-step exploitation paths that automated scanning misses entirely; helped enforce the firm's Control Procedure for red teaming AI systems.
• Every red-team finding ships with impact analysis and remediation recommendations — concrete controls partner teams use to close the gap, not just a list of vulnerabilities.
• Designing a scalable firmwide self-service red-teaming service: distributed adversarial testing on AWS with per-scanner task isolation, auto-scaling, and consolidated security report generation — MVP built on Promptfoo.
• Deployed the Network Broker handling egress and ingress for vendor-based SaaS connectivity.

JPMorganChase

AI Controls Engineering · Cybersecurity & Tech Controls

Jul 2023 — Jan 2026

Bengaluru, India

Security Engineer I

Built the AI/ML security control layer that gates every externally sourced model entering JPMC — preventing vulnerable or tampered models from reaching production systems that touch trillions in assets.

• Researched, designed, and shipped the model file scanning platform that gates every externally sourced model entering JPMC — 6,000+ models — with admission scanning at procurement plus periodic re-scanning of already-procured and home-grown artifacts, enforcing 2 Control Procedures.
• Built the scoring engine for the firm's model security inference testing tool — a risk-ranking layer over raw vulnerability-scanner output.
• Shipped an automated Kafka-based compliance pipeline covering 2,500+ models, then migrated it to an API-based architecture once the data volume no longer warranted consumer-producer overhead.
• Cut AWS billing by over $7,000 per environment (dev, test, prod) through an infrastructure revamp, lifecycle policies, and storage-class optimization.
• Spearheaded integration with the firm's CDAO Platform Registry for AI.
• Built a deserialization scanner for model files (protobuf, messagepack, h5), later replaced with a vendor-integrated solution to cut maintenance overhead.
• Designed the consolidated security report for file-scanning results — combining multiple tool outputs per model, integrated into the firm's AI Registry with a per-model security-review tab.
• Migrated deployments from Jenkins to Spinnaker; cut end-to-end test execution by ~67% and enforced secure-coding standards across the SDLC.

JPMorganChase

AI Controls Engineering · Cybersecurity & Tech Controls

Jan 2023 — Jun 2023

Bengaluru, India

Software Engineer Intern

Shipped production backend code for the ML scanning platform on day one — treated as a full engineer, not an intern. Recognized with the Super Squad Team Award for delivery quality.

• Built the core backend services and REST APIs that power the model scanning workflow — job orchestration, status tracking, and evidence collection — forming the operational backbone of what became a firm-wide ML security platform
• Brought test coverage from near-zero to 80% by authoring unit and integration test suites for every new feature, catching regressions early and enabling reliable release cadence
• Cut non-production cloud spend by over 60% by implementing Lightswitch scheduling, Janitor cleanup policies, and S3 Lifecycle tiering
• Standardized infrastructure-as-code across DEV, UAT, and PROD with Terraform — eliminating configuration drift and ensuring consistent, auditable deployments across all environments

LogicHive Solutions Pvt. Ltd.

Aug 2021 — Feb 2022

Mysuru, India

Engineering Intern

Designed and built a real-time asset tracking system from scratch — hardware to backend — giving enterprise clients live visibility into physical asset location and movement for the first time.

• Engineered the full tracking stack: Raspberry Pi + BLE beacons for device detection, MQTT for low-latency telemetry transport, and Redis for sub-millisecond location state reads — all integrated into a working PoC in 7 months
• Built the complete data pipeline — ingestion, processing, and visualization — enabling operations teams to monitor asset location, movement history, and status in real time, replacing a previously manual inventory process