Architecting Trust: How Jose Lejin P J Secures Enterprise AI

Artificial intelligence is growing fast and entering the enterprise, bringing with it a whole new set of security challenges. Static cloud environments are being replaced by dynamic, multi-agent AI ecosystems, and traditional perimeter defenses are no longer sufficient. Enterprise systems now have to guard against data leakage in large language models, dynamic application-layer attacks, and complex multi-cloud configuration gaps.

Jose Lejin P J is a Principal Member of Technical Staff at Salesforce, and has been working to solve these complicated friction points for over two decades. Jose Lejin has spent over 21 years working with large technology companies including Citrix, Cisco, Samsung, Nokia, and Motorola and built his career around identifying and fixing overlooked security vulnerabilities in high-scale platforms. Now his work helps to protect global operations dealing with billions of transactions every day.

Bridging the Gap Between Encryption and Inspection

One of the most enduring paradoxes of cybersecurity is the tension between data privacy and threat visibility. In the past, a web application firewall (WAF) needed to decrypt network traffic to be able to inspect it for malicious payloads. But this decrypt and inspect model also exposes sensitive data to the security appliance itself, making it a single point of failure, and a high-value target for attackers.

Jose Lejin solved this problem with sophisticated cryptographic techniques that were patented to allow security systems to inspect incoming requests without revealing the underlying data.

Privacy-Preserving Inspection

Rather than decrypting sensitive payloads, Jose Lejin built systems that leverage homomorphic encryption within the WAF layer. The technique enables computations and signature checks to be performed directly on the encrypted sensitive payloads, so the WAF can identify known threat signatures or anomalous patterns without decrypting the ciphertext back into plain text. The approach is designed to keep enterprise data encrypted through the inspection layer, reducing its exposure to the security appliance itself.

Contextual Threat Classification

Jose Lejin built machine-learning systems to understand intent behind network-security queries, in addition to encrypted traffic inspection. Standard firewalls are based on rigid regex-based signature matching that can easily be bypassed by slightly modified payloads.

Jose Lejin’s proprietary systems use natural language processing (NLP) and intent classification to understand the structure and semantic context of the incoming query. This allows the security layer to differentiate between a legitimate user query and a SQL injection or cross-site scripting attack, decreasing false positives in high-traffic environments.

Securing Multi-Tenant GenAI and Agentic Architectures

The arrival of generative AI platforms like Salesforce's Einstein AI and Agentforce has created a new challenge: how do you host thousands of enterprise tenants on shared AI infrastructure without the risk of data bleeding across tenants?

"Every powerful new technology is a problem-generating machine," Jose Lejin explains. "AI doesn’t retire the discipline of securing systems — it floods the world with fresh ones."

Isolating Tenants in Shared AI Spaces

Jose Lejin helped design secure boundaries in multi-tenant generative AI engines to avoid data leaks across tenants. His approach isolates tenant-specific data at the point of model inference and data retrieval. Dynamic policy-based control of access and real-time data segmentation in the system ensure that proprietary data from one tenant organization is never inadvertently shared or surfaced in outputs generated for another tenant. He has also patented methods for detecting sensitive-data leaks inside generative-AI output itself.

Defending Autonomous Agents

As enterprise AI shifts from passive chat tools to active, autonomous agents capable of executing tasks and taking action, securing communication between agents becomes critical. Jose Lejin’s work in this area spans several fronts:

• Secure Agent-to-Agent Authentication: Authenticating autonomous agents before data exchange occurs or cross-platform operations are performed.

• Post-quantum Cryptography: Cryptographic schemes designed to remain resistant to attacks from future quantum computers.

• Prompt Injection Protection: Add real-time monitoring layers to analyze agent prompts and inputs, filtering out malicious instructions that seek to hijack the agent’s behavior.

Eliminating Human Friction with Zero-Touch Security

Manual configuration is a major source of enterprise security breaches. Securing a global multi-cloud network can require updating thousands of firewalls, routers, and content delivery networks (CDNs). These updates are slow and very error prone if done manually by people. To address this, Jose Lejin developed zero-touch automation platforms that programmatically deploy and enforce network-security policies across multi-substrate cloud environments.

Automated Policy Enforcement

Such automation systems translate high-level security goals into low-level infrastructure-specific configurations. When a threat is detected or a security policy is updated, the automation engine pushes the configuration changes out globally, without human intervention.

Adaptive DDoS Mitigation

Waiting for a human operator to tune rate limits is far too slow against an active application-layer (Layer 7) distributed-denial-of-service (DDoS) attack. Jose Lejin’s architecture relies on real-time machine-learning algorithms to identify traffic anomalies, classify malicious actors, and instantly deploy targeted rate-limiting and mitigation techniques. This work was recognized with an Engineering Excellence Award for the successful defense of enterprise infrastructure under active, real-world attacks.

A Career Built on Continuous Learning

Jose Lejin’s practical engineering work is underpinned by a strong academic foundation and active involvement in the wider technical community.

He holds a Master of Technology in Computer and Information Sciences from Cochin University of Science and Technology. He has also pursued continued professional education throughout his career, including specialized programs at Stanford University in advanced computer security and design thinking, and at the Massachusetts Institute of Technology in applied data science and AI strategies.

He is a Senior Member of the IEEE and a Fellow of the Institution of Electronics and Telecommunication Engineers (IETE). Over the course of his career, he has earned multiple engineering, innovation, and security awards across the technology companies where he has worked, recognizing both his inventions and his impact on large-scale platforms. He also serves as a judge for international technology and innovation programs, evaluating early-stage ventures in technology and enterprise software.

His extensive patent portfolio with more than 50 U.S. patent filings is publicly searchable on both the USPTO Patent Public Search and Google Patents under the inventor name “Jose Lejin”. Spanning more than two decades and companies including Salesforce, Citrix, Cisco, Samsung, Nokia, and Motorola, his body of work reflects a single consistent thread: engineering security and trust into complex systems at scale, culminating in today's automated, AI-driven security architectures that protect global cloud ecosystems.