VPTReliability / Systems Start a conversation

Agent Reliability Lead · Agentic AI · Distributed Systems

I engineer systems
that have to work.

I’m Vijay Prakash Tiwari, Agent Reliability Lead at Amdocs. I engineer the controls around production AI agents and the distributed platforms beneath them—with clear architecture, measurable outcomes, and evidence behind the decisions.

Leading reliability, consistency, and performance for production agents
Vijay Prakash Tiwari, Agent Reliability Lead and systems engineer
SystemsEvidenceScale
Engineering practice
9+ years
Measured efficiency
2,000+ hours / year
Independent products
03 systems
Research record
09 published works

01 / Selected systems

Built beyond
the prototype.

Independent products that make complex engineering work legible—from graph-based agents to streaming infrastructure and career operations.

01Agent engineering platformMIT · IR v2

LangStitch

One production path across visual agent design, compilation, reusable components, and a desktop IDE.

LangStitch joins four surfaces—a typed LangGraph canvas, multi-language SDK, component marketplace, and LangTailor IDE. It exports buildable Python projects today, supports Spring AI through Maven Central, and is expanding Go and Rust targets.

  • Visual LangGraph canvas
  • IR v2 compiler & SDK
  • Marketplace + LangTailor IDE
WORKFLOW / CUSTOMER_SUPPORTVALID
TRIGGERStart
DECISIONIntent router
CONTEXTRAG
POLICYGuardrail
MODELLLM
OUTPUTResponse
02Event streamingApache 2.0

Eventore

One operational console for an event-driven estate that rarely speaks one protocol.

A modular control plane for inspecting and operating Kafka, Pulsar, RabbitMQ, MQTT, JMS, Kinesis, Google Pub/Sub, and Azure Service Bus.

CONTROL PLANE08 CONNECTORS
KafkaPulsarRabbitMQMQTTJMSKinesisGCP Pub/SubAzure SB
OBSERVEOPERATEGOVERN
03Local-first agentsBYOK

Meridian

A private career operations system that turns fragmented job search into a deliberate pipeline.

Research roles, compare job descriptions, tailor resumes, and prepare for interviews while keeping credentials and personal data under user control.

CAREER PIPELINELOCAL
  1. 01ResearchFind the right signal
  2. 02CompareMeasure role fit
  3. 03TailorShape the evidence
  4. 04PrepareClose the gaps

02 / Open research

Claims should carry
their own evidence.

My current research asks how agentic systems become inspectable, governable, and trustworthy under real operating constraints.

Evidence-Grounded Agent Routing research cover

Primary research · 2026 · Paper + implementation

Evidence-Grounded Agent Routing

Replace opaque delegation with an inspectable routing decision: evidence, eligibility, confidence, action, contract, and an auditable record.

Evaluation
42 cases
Decisions
5 outcomes
Artifact
LangGraph
Read the research overview
The 10 Pillars of Production Agent Engineering research cover

Practitioner framework · 2026 · CC BY 4.0

The 10 Pillars of Production Agent Engineering

A systems framework for the deterministic control envelope around a probabilistic model—from goals and context to execution, governance, and evaluation.

Explore the framework
Complete publication recordBrowse all nine published works

03 / Enterprise practice

Reliability, with operational consequences.

I work where model uncertainty meets production responsibility: governed agent platforms, telecom-scale billing, event-driven services, diagnostic tooling, and engineering standards.

A

Current role

Production agent reliability

Maintaining agent consistency and performance, detecting drift, protecting prompt integrity, and advancing an aSDLC playbook for reliable human–AI collaboration.

B

Agent platform

Governed AI for Digital BSS

Work spanning a foundation SDK, OpenAPI-to-MCP conversion, versioned RAG with citations, and RBAC guardrails—turning agent capabilities into operable platform controls.

C

Platform engineering

Real-time billing microservices

Java, Spring Boot, and Kafka-based integrations designed for high-throughput telecom workflows, service boundaries, and production stability.

D

Engineering leverage

Diagnostics and release automation

Built Trapper for assisted root-cause analysis and a one-click release workflow—reducing repetitive work and accelerating feedback.

04 / Capabilities

A stack shaped
by systems thinking.

Breadth is useful only when the layers connect. These are the capabilities I combine to move from architecture to reliable operation.

01

Agent reliability

Drift detection · prompt integrity · evaluation · observability · guardrails · governance · aSDLC

02

Backend engineering

Java · Spring Boot · Python · FastAPI · REST · microservices · domain modelling

03

Agent platforms

LangGraph · MCP · RAG · OpenAPI conversion · tool orchestration · RBAC · versioned knowledge

04

Event platforms

Apache Kafka · streaming architectures · integration patterns · observability · resilience

05

Platform & cloud

Kubernetes · OpenShift · Helm · CI/CD · AWS · Azure · GCP · operational tooling

05 / Experience

Built through
progressive ownership.

Nine years across integration, distributed platforms, engineering automation, and agentic systems—each role widening the operating scope.

AUG 2026 — NOW

AMDOCS · PUNE

Agent Reliability Lead

Responsible for the reliability, consistency, and performance of AI agents—including drift detection, prompt integrity, and continuous improvement of agent engineering practice.

Current
OCT 2022 — JUL 2026

AMDOCS · PUNE

Software Engineering Specialist

Designed scalable Java and Spring Boot services, Kafka event architectures, observability workflows, diagnostic tools, and release automation for enterprise telecom platforms.

DEC 2021 — OCT 2022

AMDOCS · PUNE

Software Engineer — Advanced

Built Spring Boot microservices and Kafka integrations, contributed to service design, and supported high-impact issues in distributed production environments.

JAN 2020 — NOV 2021

AMDOCS · PUNE

Software Developer — Experienced

Developed Java backend modules and REST integrations while building a foundation in telecom workflows, microservices, production support, and maintainable delivery.

JUL 2017 — DEC 2019

TIBCO SOFTWARE · PUNE

Junior Consultant

Developed enterprise integration and message-driven workflows for Macy’s, United Airlines, and Indian Railways using BusinessWorks, EMS, and BusinessEvents.

06 / Recognition

Proof, not
decoration.

2,000+

Amdocs Inspiring Innovator nominee

Global recognition for an engineering tool estimated to save more than 2,000 working hours each year.

2 / 3

Global titles at Amdocs Avishkar

Two category wins from a field of 104 teams.

716

TCS CodeVita global rank

Top 0.2% among more than 400,000 participants.

2020

Amdocs DevOps Champion

Recognized for bridging development and operations knowledge and helping improve engineering delivery practices.

07 / Profile

Engineering is the craft of making uncertainty manageable.

I work at the boundary between deterministic software systems and probabilistic agent behaviour. As Agent Reliability Lead at Amdocs, I focus on making production agents stable, observable, governed, and safe to evolve.

That practice is grounded in nine years of backend and integration engineering—and extended through open products such as LangStitch, Eventore, and Meridian, plus research on evidence-grounded delegation.

EducationB.E. Computer Engineering · A+Bharati Vidyapeeth’s College of Engineering, Lavale · SPPU · 2013—2017
Current credentialCouchbase CD215 Java DeveloperIssued April 2023 · Credential ID 158979378
Historical credentialsTIBCO BusinessEvents 5 & Spotfire 7Earned in 2018 and 2019; listed as past certifications