Staff Security Infrastructure Engineer · AI agent security, Zero Trust, cloud security · published AI-security researcher. Today: first security engineer at Parafin, which has funded $3B+ to 60,000+ small businesses and is joining Stripe (announced Sept 2026, close pending). Four peer-reviewed publications in 2026 — open access: EmbedGuard (RAG security · IJCESEN) · GenOps (AI-in-CI/CD governance · JISEM) · RuntimeGuard-AI (EU AI Act Article 14 compliance · JoCAAA) · TriArchitect (safe multi-agent Java migration · IJCISIM). Two arXiv preprints: PerceptFence (screen-share AI privacy) · The Acknowledgment Point Is the System (signed policy-decision receipts for AI audit evidence). Open source: merged adaptive-attacks probe in NVIDIA/garak (PR #1742).

name:          Neeraj Kumar Singh Beshane
location:      San Francisco Bay Area, CA
current_role:  Staff Security Infrastructure Engineer @ Parafin
previous:      [Meta Reality Labs, Raymond James, Wayfair, JPMorgan Chase, American Express]
experience:    15+ years
focus:         AI Agent Security × Zero Trust × Cloud Security
publications:  4 peer-reviewed (2026) · 2 arXiv preprints (PerceptFence, Acknowledgment Point)
writing:       28 technical articles + 13-issue "Securing the Agentic Stack" newsletter · full list on ORCID
speaking:      Conf42 DevOps 2026, Conf42 LLMs 2026, CIMA 2026 — delivered talks
open_source:   NVIDIA/garak PR #1742 · 3 merged Paperclip PRs · hermes-cursor-acp · hermes-antigravity-oauth
memberships:   Sigma Xi Full Member · IEEE Member · IICSPA Fellow · Hackathon Raptors Fellow

Four peer-reviewed papers published in 2026 — all open access — plus two arXiv preprints. Each project bundles paper + code + (where applicable) invited talk.

Cross-Layer Detection and Provenance Attestation for Adversarial Embedding Attacks in RAG Systems · IJCESEN, Vol 12 No 1, pp 589–602 · 2026-02-05

Production RAG systems are vulnerable to adversarial embeddings — poisoned vectors that look right under cosine similarity but steer downstream generation. EmbedGuard combines embedding-space anomaly scoring with provenance attestation, so an attack has to defeat both layers at once.

Key results: 94.7% detection rate on optimization-based attacks · 51 ms mean latency overhead · evaluated on a 500K-embedding production-scale system · Docker-reproducible benchmarks.

A Shared-State Multi-Agent Framework for Safe Java Code Migration · IJCISIM, Vol 18 No 18s, pp 739–751 · 2026-08-19

Multi-agent LLM systems drift when each agent holds its own private view of a codebase. TriArchitect gives them one shared, typed state — a Typed Migration Graph — and a Validator-Veto Protocol that lets any validator block an unsafe transformation before it lands.

Focus areas: software evolution · large language models · multi-agent systems · automated refactoring · legacy migration

A Governance-First Architecture for Embedding Generative AI into CI/CD Pipelines · JISEM, Vol 11(1s), pp 1518–1539 · 2026-02-15 Companion talk: Conf42 DevOps 2026 — invited speaker

A four-pillar framework for running GenAI inside production CI/CD without breaking governance: contextualization, strategic guardrails, deployment assurance, and governance audit trails.

Key results: 55.7% reduction in median deployment cycle time (52.8 → 23.4 min, p < 0.001) · zero safety policy violations · 47.2% reduction in error-budget variance · evaluated across 15,847 deployments / 127 microservices / 8 months / 3 organizations.

Scalable Tamper-Evident Accountability for High-Risk AI Systems Under the EU AI Act · JoCAAA, Vol 35 No 1, pp 1074–1090 · 2026-01-31 Companion talk: Conf42 LLMs 2026 — invited speaker · verified speaker certificate

An open-source compliance architecture for EU AI Act Article 14. Uses Groth16 zero-knowledge proofs on the bls12-381 curve to deliver cryptographically rigorous, tamper-evident audit trails inside a production latency budget.

Key results reported in the paper: 2.3–4.1% latency overhead · 62 ms witness generation · 1,389 ms total proving time at 50K-constraint scale.

Content-Mediation Architecture and Deterministic Coverage for Screen-Share AI Assistants · v0.4.0 · 2026-09-27 · with Asmita Negi

Use case: a developer or support engineer shares their screen with an AI assistant while debugging. Terminals, .env files, CI logs and admin consoles leak secrets and PII into what the assistant sees.

v0.4.0 test: 480 held-out developer-support screens rendered in Chrome, degraded, and read back by OCR, with the rules frozen before the test split ran and three screen types held out.

The cost is also measured: on the held-out screen types, 0.763 of task-relevant text survives. The preprint is on arXiv as arXiv:2609.34027 (cs.CR). It is a preprint, not peer reviewed, and is not counted among the four peer-reviewed papers.

Durable Policy-Decision Receipts for AI Audit Evidence · arXiv:2608.17176 (cs.CR) · Aug 2026 · sole author

Rebuilds RuntimeGuard-AI around the durability/latency trade-off: each policy decision gets an Ed25519-signed receipt, and receipts are chained into Merkle-epoch attestations. Preprint, not peer reviewed.

Full list of papers, articles, talks and software on ORCID 0009-0002-2125-1805.

Technical writing on LinkedIn, Dev.to and Hashnode — agentic AI security, MCP risk, RAG, Zero Trust, distributed systems. Weekly newsletter: "Securing the Agentic Stack".

Each chapter built on the last — from writing COBOL adapters to architecting Zero Trust infrastructure.

2025 ──────── PARAFIN ──────────────────────────────────────── present
 Dec          Staff Security Infrastructure Engineer
              ├─ First security engineer; built vulnerability management from zero
              │  (policy, asset tiering, priority matrix, SLA engine)
              ├─ Ran the SOC 2 Type II 2026 program end to end (12 workstreams, 86 controls)
              ├─ BC/DR from expired to exercised: live single-AZ production failover, zero customer impact
              ├─ Org-wide AWS baseline: tamper-proof CloudTrail (7-yr Object Lock), Security Hub, root MFA
              ├─ Led the company's first production TLS 1.3 migration
              ├─ OIDC + PKCE single sign-on for partner integrations
              └─ AI governance and MCP tool-access security

2024 ──────── META (REALITY LABS) ──────────────────────────── 2025 Dec
 Jun          Staff Software Engineer
              ├─ Tech lead, Reality Labs CI/CD modernization for Quest & Ray-Ban
              ├─ Build time −35% (P50), −42% (P95)
              ├─ GenAI migration tool: 95% manual effort reduction
              ├─ UBM metadata service: 450M+ requests/day, 99% fewer build-tracking failures
              ├─ Onboarding assistant: 55% oncall reduction, adopted by 3 teams
              └─ Mentored 6 senior engineers (3 promoted to Staff within 12 months)

2024 ──────── RAYMOND JAMES ────────────────────────────────── 2024 Jun
 Apr          Lead Software Development Engineer
              └─ Led a team of 8 on the client-onboarding document and e-signature platform

2022 ──────── WAYFAIR ──────────────────────────────────────── 2024 Mar
 Jan          Technical Architect
              ├─ DevSecOps platform adopted by all 15 business units
              ├─ Vulnerability management: 50M+ security checks/month
              ├─ Kubernetes policy-as-code engine (3% false positive rate)
              ├─ Unified OAuth 2.0: 125M+ API calls/week, 99.97% uptime
              └─ Mentored 20+ engineers (8 promoted to senior/staff)

2019 ──────── JPMORGAN CHASE ───────────────────────────────── 2021 Dec
 Mar          Senior Software Engineer (CIB)
              ├─ ESAP escrow portal (via Mphasis, then JPMorgan from Jul 2021)
              └─ Core platform and security lead

2011 ──────── AMERICAN EXPRESS (via TCS) ───────────────────── 2019 Mar
 Dec          Mainframe Programmer → Senior Software Developer
              ├─ International Insurance Services: monolith → microservices re-architecture
              ├─ Locations: Bangalore → Montevideo → Phoenix
              └─ This is where the journey began 🚀

Languages        Java · Python · Go · TypeScript · Groovy
Frameworks       Spring Boot · FastAPI · Node.js · React
Cloud            AWS · GCP · Kubernetes · Terraform · Docker
Security         Zero Trust · IAM · Policy-as-Code (OPA/Kyverno) · TLS/mTLS · ZK proofs (Groth16)
Data             Kafka · PostgreSQL · Cassandra · Redis · BigQuery
AI/ML            RAG Pipelines · LLM Integration · Prompt Engineering · MCP · AI Safety
DevOps           CI/CD · GitOps · ArgoCD · Datadog · PagerDuty
Patterns         Microservices · Event Sourcing · CQRS · Saga · mTLS

🐶 Dog Dad        Proud parent of Byte, a mischievous beagle
🎮 Retro Gamer    Still speed-running Contra on NES
✈️ Global Dev     Coded in Bangalore, Montevideo, Phoenix, Tampa, San Francisco
🚴 Trail Explorer System design on two wheels