Frame the real problem
Map workflows, users, constraints and the outcome worth changing.
Agastya Kommanamanchi
Forward-Deployed Engineer at Deloitte
Agastya Kommanamanchi — Forward-Deployed Engineer
I turn ambitious AI, data and cloud priorities into secure production platforms, measurable outcomes and enterprise adoption.
10+ years
5 enterprise roles
1 granted patent
Recruiter brief / At a glance
Strategy to shipped systems.
AI architecture, platform engineering and delivery leadership—connected to business outcomes.
OKR / KPI
Outcome scorecard
O1 / Reliable AI95%+task success
O2 / Faster delivery40%lower latency
O3 / Unit economics25%lower inference cost
O4 / Business value$4.5Mrecovery supported
$1.2M annual cloud savings<48h deployment cycles12% retention lift
2014—Now
Experience
- 2026—NowDeloitte
- 2022—26The Hartford
- 2021—22Accenture / PSEG
- 2020—21Cognizant
- 2014—19Reliance Jio
Multi-cloud / Multi-model
AI platform constellation
awsAWS
GGoogle Cloud
AIAnthropic
◇Databricks
✣Snowflake
Platform radar / Emerging ecosystem
Bedrock AgentCoreGemini EnterpriseGoogle AntigravityGemini 3 ProGemini Code AssistAgent Development KitClaude Enterprise
3 degrees
Education
In progressMS · Artificial IntelligenceUT Austin
2020MS · Business AnalyticsUConn
2014BTech · ECENIT Hamirpur
01 / Profile
Embedded with the problem. Accountable for the outcome.
I turn ambiguous priorities into scoped, secure systems—then stay through production rollout, adoption and measurable impact.
Forward-deployed AI • Data • Cloud
I work where customer context, hands-on engineering and executive decision-making meet.
Across regulated enterprises, I lead the path from workflow discovery and technical scoping to architecture, build, evaluation and adoption. My work spans agentic AI, identity-aware retrieval, enterprise copilots and cloud data platforms—with security, reliability and business value designed in from the start.
Make it executable
Define architecture, scope, milestones, trade-offs and delivery risk.
Own the hard middle
Prototype, integrate, evaluate and productionize with the team.
Measure what changed
Close feedback loops, prove impact and codify reusable patterns.
02 / Capabilities
The field toolkit for production AI.
Customer discovery, hands-on systems delivery and enterprise adoption—grounded in security, reliability and measurable value.
Discovery & solution shaping
Workflow decomposition, technical discovery, scoping, estimation, roadmaps and success measures
Agentic AI & LLM systems
RAG, AI agents, context engineering, tool use, prompt design, guardrails, LangGraph, ADK and MCP
Evaluation & LLMOps
Task-success evaluation, quality and safety gates, latency and cost optimization, monitoring and feedback loops
Data & retrieval platforms
Databricks, Snowflake, Spark, Airflow, retrieval pipelines, ETL/ELT and streaming
Explore the complete capability set4 additional areas
Cloud-native delivery
AWS, Google Cloud, Bedrock, Vertex AI, Kubernetes, Docker, Terraform and CI/CD
Enterprise architecture
APIs, microservices, event-driven systems, integration patterns, MLOps and reusable platform patterns
Security & AI governance
RBAC, IAM, privacy, identity-aware retrieval, responsible AI, compliance and delivery controls
Adoption & executive delivery
Stakeholder alignment, prioritization, OKRs and KPIs, rollout, change adoption, risk and executive communication
03 / Experience
A decade of building at scale.
Enterprise AI leadership backed by hands-on architecture, data engineering and product delivery.
Deloitte
Forward-Deployed Engineer, AI & Data Practice
Lead forward-deployed AI engagements from discovery and technical scoping through production launch, owning solution strategy, delivery outcomes and executive alignment.
DiscoverySolution architectureProduction rolloutExecutive alignment
- Translate ambiguous priorities into AI product roadmaps, scopes, staffing models and executable delivery plans.
- Architect secure AI and data platforms across AWS, Google Cloud, Databricks and Snowflake.
- Guide CIO-level stakeholders through prioritization, architecture trade-offs, adoption and delivery risk.
The Hartford
Senior Staff Software Engineer
Owned prototype-to-production delivery for GenAI, enterprise copilot and agentic engineering initiatives across insurance workflows.
Prototype to productionAI evaluationPlatform adoptionReusable patterns
- Delivered secure GenAI underwriting assistants and identity-aware RAG pipelines.
- Scaled enterprise Copilot and GitHub-based agentic remediation workflows.
- Reached 95%+ task success while reducing latency 40% and inference cost 25%.
Earlier experienceAccenture / PSEG • Cognizant • Reliance Jio 2014—2022
Accenture • Client: PSEG
AI Solutions & Data Products - Senior Consultant
Built an AWS Customer 360 data platform and operationalized predictive models for utility programs.
Data integrationML operationsCI/CDBusiness outcomes
- Unified CRM, billing, meter and operational data through real-time and batch pipelines.
- Improved retention by 12% and supported $4.5M in revenue recovery.
- Reduced deployment cycles from weeks to under 48 hours with CI/CD and IAM controls.
Cognizant Technology Solutions
Senior Cloud Architect & Product Developer
Led development of a multi-cloud FinOps platform used by more than 200 engineers.
Multi-cloudFinOpsPlatform engineeringModernization
- Built billing pipelines, anomaly detection and cloud cost visibility across AWS and GCP.
- Modernized legacy workloads with Kafka, Kubernetes, Snowflake and Databricks.
- Reduced annual cloud costs by more than $1.2M.
Reliance Jio
Deputy Manager - Data Science and Product Analytics
Built large-scale telemetry, recommendation and product analytics systems.
Distributed dataRecommendationsExperimentationProduct analytics
- Processed more than 10TB of daily telecom telemetry with Hadoop and Spark.
- Served behavioral models to 50M+ users, improving session duration 25%.
- Improved click-through rate 14% through data-driven interventions.
04 / Selected Systems
Built for production, measured by outcomes.
Public case studies drawn directly from professional experience.
01Enterprise GenAI
Underwriting intelligence, grounded and secure.
Led the design and delivery of GenAI underwriting assistants, translating document-intensive workflows into agentic systems with identity-aware retrieval, evaluation and production controls.
95%+ task success40% lower latency25% lower inference cost
AWS BedrockVertex AIRAGKubernetesRBAC
02Agentic Engineering
Repository intelligence for remediation at scale.
Delivered GitHub-based workflows that analyze repositories, classify technical debt and orchestrate remediation through APIs and automation.
GitHubAgentsAPIsCI/CD
03Data & ML
Customer 360 for utility intelligence.
Unified customer, meter and operational data on AWS, then embedded predictive outputs into reporting and business workflows.
12% better retention$4.5M recovery supported
S3GlueKinesisSageMakerTableau
04Cloud Platforms
Multi-cloud FinOps for engineering teams.
Built cost visibility, anomaly detection and automated infrastructure controls across AWS and GCP.
200+ engineers$1.2M annual savings
AWSGCPKafkaTerraform
05Large-Scale Analytics
Behavioral intelligence for 50M+ users.
Led recommendation, telemetry and experimentation systems operating on more than 10TB of daily data - turning behavioral signals into measurable product outcomes.
50M+ users25% longer sessions14% higher CTR
HadoopSparkRecommendationsA/B Testing
Granted intellectual property / Jan 6, 2026
System and method for a generative artificial intelligence model gateway
A secure enterprise AI gateway that evaluates prompts for sensitive information, governs access to approved language models and supports managed interactions with model outputs.
Sensitive-data controlsModel access governanceManaged AI interactions
- Publication
- US12518057B2
- Filed
- Oct 18, 2023
- Assignee
- Hartford Fire Insurance Co.
Inventors: Senthilkumar Gnanasekaran, Agastya Kommanamanchi, Shrujan Jyotindrabhai Mistry and Renoi Thomas.
View the granted patent
05 / Credentials
Verified, current and continuously learning.
12 current public credentials, two professional certifications and completed Anthropic training—organized by provider for quick review.
Credential archive3 expired credentials
06 / Open Source
A public lab for AI systems.
Five original projects are featured first. Forked samples and ecosystem explorations remain available through the archive filters.
07 / Insights
Writing and visual storytelling.
Technical perspectives on production AI alongside a decade-long practice in data visualization.
Technical writing
All articles ↗Tableau Public
10 published visualizations across geospatial, statistical and business analytics.
08 / Foundation
Education and industry engagement.
Education
Master of Science, Artificial Intelligence
The University of Texas at Austin
Master of Science, Business Analytics and Project Management
University of Connecticut
Bachelor of Technology, Electronics and Communications
National Institute of Technology Hamirpur
Conferences & training
Google Cloud Next
Las Vegas
Deloitte Forward-Deployed AI Academy & Training
Industry engagement
AWS re:Invent
Las Vegas
HashiCorp Conference
San Francisco
10 / Introduction video
Why I love forward-deployed engineering.
A 43-second view of how I connect customer context, hands-on engineering and measurable enterprise outcomes.