Varun Reddy

Houston, TX · Applied AI Architect

One résumé. Two lenses.

18+ years turning complex AI and data problems into platforms that save money, reduce risk, and ship. I translate between the boardroom and the build — so the right thing gets built, and everyone understands why.

Get in touchlinkedin.com/in/reddyvarunSwitch the lens above — the page rewrites itself for your audience.

$15M+

in measurable savings delivered

Across enterprise platforms over a decade — anchored by a vendor/expenditure app that saved $2M+ in its first six months alone.

95%

less manual data handling

Master-data management improvements across multiple applications and platforms.

70%

fewer manual HSE tasks

Redesigned compliance data models for consistent availability and reporting.

50%

higher data reliability

Snowflake data-quality plugin detecting missing and duplicate records.

The short version

Senior architect and technical leader who has led cross-functional teams across Oil & Gas, HSE, and HR/HCM. I am the person who can sit with an executive to frame the business case and then sit with engineers to design the system that delivers it — closing the gap that usually stalls enterprise AI.

Experience

Apr 2026 — Present

AI Architect

Wissen Infotech

Client: EOG Resources

Houston, TX

  • Architect a multi-agent AI platform that lets business teams safely build their own AI skills on a governed foundation.
  • Defined the technical standards, performance rules, and human-in-the-loop patterns that every skill builder across the company must follow.
  • Brought multimodal AI to real operations — extracting structured data from documents and analyzing live facility camera feeds.

Jan 2014 — Apr 2026

Development Architect

Wissen Infotech

Client: ConocoPhillips / Marathon Oil

Houston, TX

  • Built a full-stack vendor/expenditure monitoring app that saved more than $2M within six months of go-live.
  • Cut manual data handling by up to 95% across applications, enabling faster, more accurate reporting.
  • Reduced manual HSE compliance tasks by 70% and improved data reliability by 50% through a Snowflake data-quality system.
  • Led teams of data engineers and developers across cloud and on-prem platforms, aligning delivery with business goals.

Oct 2015 — Apr 2017

Development Consultant

Wissen Infotech

Client: SAP America

Newtown Square, PA

  • Shipped a cloud gamification service into SAP SuccessFactors that became a paid extension offering, boosting learning engagement.
  • Partnered directly with SAP's Cloud Platform team to shape the product's design.

Jul 2008 — Jan 2014

Senior Developer (T3)

SAP Labs India

Bengaluru, India

  • Led feature development across multiple product versions of SAP's talent-management suite, coordinating cross-functional teams and timelines.
  • Hardened data security across HCM and FI modules to meet global compliance standards.

May 2007 — Jul 2008

Programmer Analyst

Cognizant Technology Solutions

Bengaluru, India

  • Built UI for Intuit QuickBooks, improving functionality and user experience for a leading small-business accounting product.

Safety & Evaluation

Reliable, interpretable, steerable — built in, not bolted on.

The hard part of enterprise AI isn't the demo — it's trust. Here's how I make models safe to put in front of a business.

Guardrails, in and out

Prompt-injection and content-filtering safeguards on both pre- and post-processing — deterministic (regex, file-type, keyword) and model-based (NLP, OpenAI) — to block unsafe or biased outputs.

Evaluation engineering

Offline evals (BLEU, ROUGE, prompt linting), online evals (A/B and cohort testing), and live user-feedback loops. Golden eval suites gate prompt changes before they ship.

Controlling model behavior

Applied tokenization, context-window management, temperature/top-p tuning, and system-prompt design to control soundness, determinism, and creativity — mitigating context loss, truncation, hallucination, and prompt drift.

Steerable by design

Platform-level skills constitution and architectural patterns (response formatting, human-in-the-loop, input requirements) that keep business-built skills safe and consistent.

Capabilities

AI / ML

  • LLMs (OpenAI, Anthropic Claude, Gemini)
  • RAG
  • Prompt / Context Engineering
  • Eval Engineering
  • MCP
  • Agents / A2A
  • LangChain
  • LangGraph
  • CrewAI
  • Langfuse
  • Vector DBs (Neo4j, FAISS)
  • OCR
  • Fine-tuning / Inference Optimization

Architecture & Design

  • Microservices
  • Event-Driven Architecture
  • API-first vs MCP design
  • Cost Optimization
  • Governance

Data & Platform

  • Snowflake
  • Apache Airflow
  • NiFi
  • Kafka / Kafka Connect
  • Docker
  • Kubernetes (EKS)
  • AWS (S3, EC2, MSK)
  • Data Quality & Governance

Languages

  • Python
  • Node.js
  • TypeScript / JavaScript
  • SQL
  • CypherQL
  • ABAP
  • Bash

Databases & Storage

  • Snowflake
  • MongoDB
  • Redis
  • SAP HANA
  • Neo4j
  • PostgreSQL
  • OpenText

Leadership & Delivery

  • Cross-functional Team Leadership
  • Agile / SCRUM
  • Enterprise Data Strategy
  • Build vs Buy decisions
  • Architecture review
  • Mentorship

Education

PG in Artificial Intelligence / Machine Learning

UT Austin

2020 – 2021

B.E. Electrical & Electronics Engineering

College of Engineering, JNTU Hyderabad

2003 – 2007

Let's talk

I help enterprises turn Claude and LLMs into systems they can trust.

varun.paradarami@gmail.comLinkedIn