Lab file · Unit SM-26 · Sunnyvale, California AI Engineer · Sunnyvale, California

Dr. Gero built androids. I build agents.

I make LLMs do real work in production.

Hi, I'm Srikar. Right now I build agents that investigate CI failures and automate the SDLC for Cisco's NX-OS teams. Before that: an agentic procurement platform, CLIP trained from scratch, and a lot of computer vision. 7+ years of shipping AI that people actually use.

GERO-SUPERCOMPUTER :: DIAGNOSTICProfile
Srikar Madarapu
  • > run diagnostic --unit srikar
  • UNIT Srikar Madarapu
  • CLASS AI Engineer · Agentic · CV · Multimodal
  • CURRENT FDSE, Applied AI @ Cisco
  • RUNTIME 7+ years in production AI
  • EFFICIENCY −85% manual ops · −50% test time
  • COST / LATENCY −60% / −40%
  • GUARDRAILS ON · HUMAN-IN-THE-LOOP: ON
  • STATUS: READY FOR ACTIVATIONOpen to new roles

85 %

Manual operations removed

AI procurement platform

60 %

Lower AI workflow cost

Semantic log filtering

40 %

Lower AI workflow latency

Pattern classification

50 %

Faster regression runs

4h → 2h on NX-OS

00 / Choose a path

Hop on a cloud.Explore

UNIT 16 · GUIDE Grab the Flying Nimbus — Experience for the route so far, Projects for what shipped.

01 / Where I've worked

ExperienceExperience

UNIT 16 · GUIDE Nine stops on the route — roles as forms, college as training places. Newest first.

  1. Mar 2026 – Present UI ● CURRENT Forward Deployed Solutions Engineer – Applied AI Cisco (Contract) · Sunnyvale, CA Agentic SDLC automation
    • Owned discovery, prototyping, and rollout of agentic automation for Nexus NX-OS engineering teams in Data Center Networking.
    • Built CI investigation and remediation agents correlating Jenkins logs, test results, and code changes — separating product defects, flaky tests, and infra failures, with human-approved fixes.
    • Reduced AI workflow latency by 40% and cost by 60% with semantic log filtering and pattern classification, reserving LLM reasoning for hard cases.
    • Cut regression execution time by 50% (4h → 2h) through regression analysis and test workflow optimization.
    • Built shared agent infrastructure and a unified RAG system powering Webex bots, with citations, metadata filtering, and retrieval evaluation.
    • Integrated pyATS/Genie checks across NX-OS, IOS XE, and IOS XR to validate device state before and after approved changes.
    • Deployed and scaled Langfuse and built agent evaluation harnesses; contributed to AI-assisted code review spanning triage, fixes, testing, and merge.
  2. Jan 2025 – Mar 2026 Blue Full-Stack AI Engineer Right Pro Inc · Sunnyvale, CA Autonomous procurement
    • Led end-to-end development of an AI-powered procurement platform, reducing manual operations by 85%.
    • Architected agentic systems for multistep planning and execution with structured tool calling, persistent memory, secure action schemas, and policy guardrails.
    • Implemented MCP servers and a decorator-based tool registry for automatic API discovery and agent–tool integration.
    • Built secure email and document agents with Microsoft Graph API and Azure AI Document Intelligence on a multi-tenant Azure Blob layer.
    • Shipped via Docker, Azure DevOps, and Argo CD with logging, metrics, and tracing for production SLAs.
    • Drove 0→1 product development and owned system design, architecture, and production rollout.
  3. Aug 2023 – Nov 2024 SSJ4 Computer Vision Research Assistant · Graduate TA MedIX Lab, DePaul University · Chicago, IL Multimodal research
    • Trained OpenAI CLIP from scratch for semantic image search with FAISS tuned for sub-second retrieval.
    • Fine-tuned networks for lung-cancer CT multi-classification, improving diagnostic accuracy by 8%.
    • Cut model training time by 75% with MLflow tracking and hyperparameter optimization.
    • Compared Vision Transformers vs CNNs for limited medical datasets.
    • Supported 60+ students in data science, ML, and image processing as a Graduate TA.
  4. Jan 2023 – Nov 2024 Kai ● TRAINING Master's in Data Science DePaul University · Chicago, IL King Kai's planet · Gravity training
    • Completed a Master's in Data Science with a focus on computer vision and multimodal learning.
    • Built research projects spanning CLIP, medical CT classification, and retrieval systems.
    • Supported 60+ students as a Graduate TA across data science, ML, and image processing.
  5. Sep 2019 – Dec 2022 SSJ3 AI & Visual Data Analytics Founder SRP · Bangalore, India Data factory
    • Built and grew a visual media startup: 20+ projects with 6+ partner companies across India.
    • Annotated image/video datasets in CVAT and built ETL pipelines for ML-ready annotations.
    • Built aerial and 360° image/video data warehouses for AI-driven projects.
    • Ran analytics and A/B testing that drove a 20% lift in conversions.
  6. May 2019 – Aug 2019 SSJ2 Video Analytics Engineer Origa Labs · Bangalore, India Real-time vision
    • Designed real-time motion detection for human activity monitoring, improving machinery safety alerts by 35%.
    • Engineered multi-object tracking with Kalman filters and Deep SORT to keep identity through occlusions.
    • Achieved 3× faster edge inference with OpenVINO quantization and pruning.
  7. May 2018 – Sep 2018 SSJ Computer Vision Engineer Inferigence Quotient LLP · Bangalore, India Spatial awareness
    • Benchmarked open-source SLAM systems with camera and IMU data.
    • Improved indoor localization accuracy by 5% and calibrated cameras for 3D mapping.
  8. Jul 2017 – Dec 2017 Base Research Intern DRDO CAIR · Bangalore, India First boot
    • Implemented CNN-based monocular depth prediction (78% accuracy).
    • Integrated depth prediction with SLAM for real-time robotic navigation.
  9. Jul 2014 – May 2018 Roshi ● TRAINING Bachelor's in Computer Science NIIT University · Neemrana, India Turtle School · Turtle School basics
    • Built the CS foundation: algorithms, systems, and the first computer-vision projects.
    • Interned at DRDO CAIR during senior year — depth prediction and SLAM for robotics.
    • Graduated ready to ship real vision systems, not just coursework demos.

02 / What I've built

ProjectsProjects

UNIT 16 · GUIDE Things that shipped to production.

Cisco

CI Investigator

Looks at a failed CI run — Jenkins logs, test results, the diff — and works out whether it's a real bug, a flaky test, or broken infra. Then proposes a fix and waits for a human to approve it.

  • LLM reasoning
  • Semantic log filtering
  • Langfuse traces
Guardrails
ON
Human-in-loop
ON

Cisco

Unified RAG Core

One retrieval system behind Webex bots and internal tools. Answers come with citations, and retrieval quality is measured, not guessed.

  • RAG
  • Metadata filtering
  • Retrieval evaluation
Guardrails
ON
Human-in-loop
AUTO

Right Pro

Procurement Agent

Plans and runs multistep procurement tasks on its own. Took about 85% of the manual work out of the process.

  • Tool calling
  • Persistent memory
  • Policy guardrails
Guardrails
ON
Human-in-loop
ON

Right Pro

Inbox & Docs Agent

Reads incoming email and documents, pulls out what matters, and files it away — securely, per tenant.

  • MS Graph API
  • Azure Doc Intelligence
  • MCP tool registry
Guardrails
ON
Human-in-loop
ON

DePaul MedIX

Semantic Seeker

Search images with text or another image. CLIP trained from scratch, FAISS for sub-second lookup.

  • CLIP
  • FAISS
  • PyTorch
Guardrails
ON
Human-in-loop
AUTO

Origa Labs

Safety Tracker

Tracks people around industrial machines in real time, and doesn't lose them when they walk behind something.

  • Deep SORT
  • Kalman filter
  • OpenVINO edge
Guardrails
ON
Human-in-loop
AUTO

03 / Installed modules

Core systemsSkills

UNIT 16 · GUIDE The toolbox. Mostly Python and PyTorch, plus whatever the problem needs.

Agentic & LLM Systems

18 loaded
  • LLMs
  • RAG
  • LangChain
  • LangGraph
  • LangSmith
  • Langfuse
  • LiteLLM
  • TruLens
  • CrewAI
  • OpenAI Agent Builder
  • ChatKit
  • Prompt Engineering
  • FAISS
  • Pinecone
  • Milvus
  • Temporal
  • Hugging Face
  • Copilot

Programming & Frameworks

9 loaded
  • Python
  • TypeScript
  • Node.js
  • FastAPI
  • React
  • PostgreSQL
  • MongoDB
  • Linux
  • REST / GraphQL

Machine Learning & AI

14 loaded
  • PyTorch
  • Transformers
  • CLIP
  • Vision Transformers
  • CNNs
  • YOLO
  • Segmentation
  • SLAM
  • Few-shot Learning
  • Transfer Learning
  • Scikit-learn
  • Pandas
  • NumPy
  • JAX

MLOps & Deployment

11 loaded
  • Docker
  • Kubernetes
  • CI/CD
  • Jenkins
  • Airflow
  • Argo CD
  • MLflow
  • Weights & Biases
  • Metaflow
  • ONNX
  • Model Optimization

Cloud & Enterprise

11 loaded
  • AWS S3
  • SageMaker
  • Lambda
  • EC2
  • Bedrock
  • Kafka
  • Azure AI Foundry
  • Azure Doc Intelligence
  • Azure Blob
  • Snowflake
  • Databricks

04 / Unit profile

Engineer specificationAbout me

UNIT 16 · GUIDE The person behind the units.

LAB ID · SM-26Profile
Srikar Madarapu

Srikar Madarapu

Role
Forward Deployed Solutions Engineer – Applied AI, Cisco
Based in
Sunnyvale, California
Runtime
7+ years

I've spent 7+ years building AI systems that run in production — first computer vision, now LLM agents. I like taking a messy real-world workflow, figuring out which parts actually need a model, and building the whole thing end to end: retrieval, tools, evals, deployment, and the unglamorous reliability work that makes it stick.

OFF-DUTY MODE:Off the clock: Photography · Road trips · Board & card games

More about me →

Training dataEducation

  • Master's in Data Science DePaul University Jan 2023 – Nov 2024
  • Bachelor's in Computer Science NIIT University Jul 2014 – May 2018

Certified components · 07 Certifications

  • AWS Certified Generative AI Developer
  • AWS Certified AI Practitioner
  • Oracle Cloud Generative AI Professional
  • AI Code Review — DeepLearning.AI
  • AI Agents for Image and Video Generation — DeepLearning.AI
  • Long-Term Agentic Memory with LangGraph
  • Building Ambient Agents with LangGraph

05 / Field logs

Lab notesRecent writing

All field logs →All posts →

06 / Activation

Want to build something?Get in touch

$ activate --unit srikar --mission "your-team"

If you have a workflow that should be an agent, or an agent that keeps breaking, I'd like to hear about it.

Powering up

POWER LEVEL 0005

Ultra Instinct Opening experience

Sprite credits

Character art on this site comes from the community Dragon Ball API (transparent WebP renders) — Goku's forms for career stops, plus Master Roshi and King Kai for college training places. Dragon Ball, Dragon Ball Z, and related characters are trademarks of Toei Animation / Bird Studio / Shueisha. This portfolio is a fan tribute and is not affiliated with them.

Forms

Training places

Inspired by interactive portfolios like yash-kumar.link.