Fifteen years of production systems, now agentic AI that enterprises can audit.

Anil Jonnala

Most enterprise AI initiatives stall because they treat LLMs as chat novelties rather than untrusted, non-deterministic components in a distributed system. I bridge that divide.

Over 15 years engineering mission-critical backends across tier-1 telecommunications and financial institutions, I have designed high-concurrency event pipelines, decomposed legacy monoliths into resilient Kubernetes microservices, and built enterprise observability fabrics that cut MTTR by ~25%.

Today, I architect production-grade multi-agent platforms—such as Inforsight and ObservAI—grounded in LangGraph state machines, GraphRAG, and hybrid retrieval. Autonomous agents are governed by deterministic guardrails, automated RAGAS evaluations, and verifiable proof chains, ensuring they remain resilient and safe for regulated enterprise deployment.

Deterministic AI Safety

Hard confidence bounds, deliberate abstention protocols, and human-in-the-loop gates preventing unverified autonomous actions.

Distributed Systems Scale

Event streaming on Kafka, container orchestration on AWS EKS, and high-throughput non-blocking asynchronous APIs.

Full-Stack Observability

Execution-graph tracing, OpenTelemetry instrumentation, and synthetic health monitoring cutting incident resolution times.

Live Systems

Inforsight

Personal Build · Policy Intelligence & Multi-Agent Governance

A personal build demonstrating verifiable enterprise AI. Stateful agents score risk and cite verifiable evidence under strict determinism: they can never bypass human review or execute unauthorized actions.

Personal build — fictional enterprise inputs, real backend responses. Select a scenario to inspect the execution trace:

01 A Late Payment

Full lifecycle: event ingestion → modeled risk → eligibility → recorded human decision.

02 Missing Safety Evidence Agent Abstains

Deliberate abstention: unknown legal-hold facts halt action and prevent hallucinated approvals.

03 A Known Legal Hold Safety Constraint

Deterministic safety enforcement: active legal hold blocks action and defers to human reviewer.

Scenario 01 · A Late Payment Deterministic Review Gate

Event Ingestion: kafka:billing.payment.delinquent

Customer Tier: Gold · Days Past Due: 34 · Outstanding: $1,420

Policy Evaluation: HardshipGraceRule_v2

Risk Score: 0.74 · Confidence: 0.96 · Evidence: Prior payment streak intact

Human Authority Gate: YIELD_TO_OPERATOR

Deterministic safety rule triggers manual review for balance > $1,000

Immutable Audit Ledger: RECORDED

SHA-256: 0x7f4e91...b820 · State: Pending human signoff

Java Control Plane → Kafka Event Bus → Python Policy Agent → Immutable Audit Ledger

Launch Live Scenario Simulator

Capabilities

Agent orchestration LangGraph, LangChain, OpenAI and Claude APIs, Model Context Protocol, tool calling, multi-agent workflows, streaming inference, memory and checkpointing
Agent observability Execution-graph tracing, tool-call tracking, OpenTelemetry, token and latency monitoring, RAGAS evaluation, chain-of-evidence tracing
Retrieval Hybrid RAG and GraphRAG, vector search, embeddings, chunking strategies, LlamaIndex, pgvector, Neo4j, blast-radius analysis
LLMOps and governance Databricks AI Gateway, Delta Lake, guardrails, model cards, PII redaction, insert-only audit logs, model risk management
Cloud and infrastructure AWS (EKS, Lambda, Bedrock, S3, SQS), Kubernetes, Docker, Helm, Terraform, GitHub Actions, HashiCorp Vault, OIDC and RBAC
Backend and integration Python and FastAPI, Java and Spring Boot, React, Node.js, REST and WebSockets, Kafka, Postgres, ServiceNow and Databricks integrations

Certifications

Certified Kubernetes Application Developer (CKAD)
AWS Certified Solutions Architect, Associate

Education

B.Tech, Computer Science and Engineering, JNTU Kakinada, 2010

Dec 2010 to Aug 2015

Reliability

Systems Engineer

Telecom Scale British Telecom · Mission-Critical Infrastructure (via TCS)

99.999%verified SLA uptime sustained across international telecom backends

Eliminated single points of failure across tier-1 UK telecom infrastructure through service virtualization and fault-tolerant routing.
Automated business-critical failover workflows with Unix and Java, establishing real-time WebLogic monitoring across distributed environments.

Sep 2015 to Jul 2018

Throughput

Java Developer

Financial Services New York Life · Core Policy & Transactional Data (via TCS)

+20%transaction throughput on enterprise data pipelines

Architected high-throughput REST APIs and non-blocking integration middleware in Java and Spring Boot for core insurance data pipelines.
Pioneered containerized microservice patterns and established the organization's initial cloud-native CI/CD delivery pipelines.

Aug 2018 to Sep 2022

Modernization

Senior Java Developer · Cloud Modernization

Cloud Architecture New York Life · Enterprise Cloud Migration (via TCS)

Zerounplanned downtime during multi-year monolith decomposition

Spearheaded multi-year architectural decomposition of core monolithic insurance platforms into highly available microservices on Amazon EKS with Helm.
Architected resilient event-driven streaming pipelines using Apache Kafka, AWS Lambda, SQS, and S3 for high-throughput asynchronous workloads.

Oct 2022 to Mar 2024

Observability

Lead Software Engineer · Enterprise Observability & SRE

Observability & SRE New York Life · Telemetry Fabric & Incident Command (via TCS)

~25%reduction in mean time to detect and resolve (MTTD / MTTR)

Served as technical lead for enterprise observability fabric, standardizing telemetry across traces, metrics, and logs for distributed microservices.
Engineered an exception-first command center using custom New Relic NR1 applications tracking synthetic monitor health across tier-1 systems.

Apr 2024 to now

Intelligence

Lead Software Engineer · AI & Data Systems

Autonomous Systems Enterprise AI Architecture · New York Life (via TCS)

Zero unverifiedautonomous actions permitted across production decision workflows

Architected multi-agent incident intelligence platforms in Python and LangGraph, orchestrating stateful agents with deterministic confidence gates.
Built hybrid retrieval over pgvector and Neo4j knowledge graphs on Databricks Delta Lake and AWS Bedrock, backed by automated RAGAS evaluations and immutable audit logs.

Let’s talk about what you’re building.

aniljonnala.89@gmail.com