Full Stack AI Engineer & Developer · Open to Full-Time & Consulting
Hands-on coder for 30 years. Polyglot. Cloud-native. Now a hands-on agentic coder building production AI — steering Claude Code, GitHub Copilot, and frontier LLMs as collaborative peers, never prompt generators. The constant is the problem; the variable is the stack. I have solved the same enduring domains across every technology generation: Perl → Java → MQ/JMS → Kafka → Agentic AI. I make decisions and ship code.
Surendra Kashyap · Lombard, Illinois · Open to Austin · Chicago · Remote · Full-Time or Contract
Featured Use Cases
A production multi-agent system that transforms unstructured client communications into executable financial workflows. Animated walkthrough + SMART STAR Q&A.
Vector embeddings, KNN retrieval, and LLMs over the QQQ prospectus — the foundation I built before agentic AI. Chunking, embedding, and live Q&A simulation.
End-to-end A2A orchestration: supervisor-worker delegation, validator → initiator pipelines, RAG with Vector DBs (Kendra · Pinecone · Weaviate), Graph DBs (Neo4j · AWS Neptune), W3C distributed tracing. GPT-4o · Claude Opus · Gemini · AWS Bedrock. Transferable to any document-heavy regulated domain — financial services · insurance · healthcare.
Take an existing AI POC to production: prompt engineering, multi-pass continuation, structured JSON extraction, ethical-AI safeguards, feature-flag-first delivery, 170+ E2E regression suites designed for LLM non-determinism, PII redaction, and compliance-aware prompts (SOX · GLBA · HIPAA).
Help your engineering team adopt Claude Code, GitHub Copilot, and frontier LLMs as collaborative peers — architecturally directive, never prompt-dependent. I shift teams from Generative AI to Action AI: agents that comprehend, decide, validate, and execute.
25-year messaging lineage: IBM MQ → JMS → Mule ESB → Kafka. Event-driven architecture is not a pattern I learned — it is how I think. Cloud-native on AWS (EKS · ECS · Bedrock · Step Functions · MSK Kafka) with Terraform, Spinnaker, and Harness. Polyglot: Python · Java · Node.js.
From problem framing to production operating model. I don't hand off at the demo.
Code is the deliverable. I make decisions, write code, and stay close to the outcome.
The stack changes every five years. The problems don't. I re-learn each generation hands-on, refactor my own thinking, and build systems that are evolvable — never frozen.
Success is measured in business results: cognitive burden removed, transactions completed, time given back.
Compliance-aware prompts, deterministic validators, audit trails. Trust is engineered, not bolted on.
2–4 weeks
Discovery + working POC. I frame the problem, prototype the agent architecture, and prove or disprove feasibility with real data.
Outcome: A clear go/no-go and a defensible plan.
8–16 weeks
Whiteboard to production system. I design, code, harden, and ship the agentic system end-to-end.
Outcome: A deployed system handling real traffic with full observability.
3–6 months
Hands-on architect inside your team. I own the AI architecture, code daily alongside your engineers, and hand off a system your team can extend.
Outcome: Capability transfer, not dependency.
Ongoing
AI-augmented engineering practice for an existing team. I help your engineers adopt agentic coding tools and build their own production AI muscle.
Outcome: Team-level uplift.
years of hands-on software engineering across financial services, platform engineering, and enterprise systems — J.P. Morgan Chase, TD Ameritrade, Morgan Stanley · Invesco, GXS · GE, Saint Thomas Health.
years in messaging, event-driven architecture, streaming, and ETL — IBM MQ → JMS → ActiveMQ → Mule ESB → Informatica PowerCenter → Kafka. Every system I ship is decoupled, replayable, and observable.
years in production Generative AI and LLM integration — GPT-4o, Claude Opus, Gemini, Azure OpenAI, AWS Bedrock, Amazon SageMaker.
years as sole architect, coder, and operator of a multi-agent Supervisory AI Agent at J.P. Morgan Chase — ingests unstructured client emails, decomposes intent, delegates to skill-based sub-agents, and initiates real financial transactions across global asset & wealth management. 170+ E2E regression suite designed for LLM non-determinism.
I earn trust by learning alongside the people I work with — not lecturing from a podium. I started pursuing formal credentials in 1997 and I have never stopped. These aren't badges to put on a wall; they are proof that I hold myself to the same standard I ask of every team I coach. Lead by example. Learn in public. Grow together.
Mar 2026
Anthropic
Certified in production Agentic AI engineering with Claude — the same toolchain I use on every client engagement today.
May 2025
Cornell University · eCornell
Strategic framing for AI adoption at the enterprise level — bridges the gap between what models can do and what organizations actually need.
Nov 2024 · Active to 2027
Amazon Web Services
Cloud-native architecture validated at the source. Every system I design is built on this foundation.
Jan 2024 · Active to 2027
Amazon Web Services
ML at the infrastructure level — SageMaker pipelines to model deployment on AWS. Specialty-level validation.
May 2024 · eCornell · 3 months
Cornell University
Rigorous, condensed ML program from a top-tier institution — the academic foundation behind the AI engineering I practice daily.
Jan 2023
HashiCorp
Infrastructure as code is non-negotiable for any modern platform. I certify what I practice.
Jan 2020
Cloud Native Computing Foundation
Container orchestration at the application level — the operational discipline behind every cloud-native system I have shipped.
Jul 2020
Amazon Web Services
Developer-level AWS validation. Built the muscle memory for AWS SDKs, Lambda, and event-driven architecture.
Sep 2001
Sun Microsystems
Enterprise Java from the source — where my discipline for production-grade, rigorous engineering began.
2001 – 2003
University of Phoenix · GPA 3.89
Second graduate degree. Systems thinking at enterprise scale, earned while working full-time — proof that the learning never stops.
1991 – 1994
NIT Raipur · Application Development
Graduate-level CS from a premier Indian institution — the engineering foundation I have built everything on since.
1986 – 1989
Pandit Ravishankar Shukla University
Sciences grounding before engineering — analytical thinking, first principles, and a comfort with complexity that has never left me.
Lead AI Engineer · Hands-On Agentic Coder · End-to-End Owner
I have spent 30 years as a practitioner, never a passenger. Polyglot across Python, Java, Node.js, and modern frontend. I run an informal AI learning community — because the fastest way to stay sharp is to teach, debug, and build alongside people who are still growing. That mutual energy is what I bring to every engagement. Based in Lombard, Illinois.
English (Fluent) · Hindi (Native) · Spanish (Elementary)
Hiring a Full Stack AI Engineer or Staff-level AI Developer? I'm actively looking — full-time or contract, on-site or remote.
Also available for consulting: production AI agents, platform modernization, or team coaching in agentic engineering.