g2i
Evaluation engineeringG2i
AI evaluation & benchmark engineering
Three areas of freelance work spanning task design, comparative model evaluation and the maintainability of generated code.
Read the work overview NOTES FROM AN AI ENGINEER
I build productionAI systems.
I’m Uday, an AI engineer working across enterprise retrieval, agent workflows and Python services. These are the systems I’ve built, the problems behind them and the lessons along the way.
a little less magic. a lot more evidence.
FROM IDEA TO PRODUCTIONPython & backend systemsRetrieval & agentsBenchmarks & evaluation
01 / PROFESSIONAL EXPERIENCE
AI systems,
in production.
Professional experience across enterprise AI, automation, recommendations and applied machine learning.
- Aug 2025 – Present
Netskope
Senior Developer · AI & Automation
Production RAG and agent systems
Explore 1 project & architecture notes - Jul 2022 – Aug 2025
ServiceNow
Machine Learning Engineer
Enterprise AI and incident automation
Explore 1 project & architecture notes - Jun 2021 – Jun 2022
SWYM
Software Engineer
Recommendations and commerce analytics
Explore 2 projects & architecture notes - May 2020 – May 2021
Dhan AI
Machine Learning Engineer
Virtual assistants and healthcare NLP
Explore 2 projects & architecture notes - Oct 2018 – May 2020
Synopsys
Technical Engineer · Machine Learning
Applied ML and data systems
Explore 2 projects & architecture notes
02 / FREELANCE EXPERIENCE
Freelance & evaluation work.
Selected engagements in coding benchmarks, model evaluation, task design and computer-use data.
eq
Task authoringExpertquery
Software-engineering task authoring
Realistic features, fixes and enhancements built around behavioral requirements, reference implementations and held-out tests in reproducible environments.
Read the work overviewa
Comparative evaluationAlignerr
Long-horizon model evaluation
Compare two models on extended build tasks, writing rubrics and rationales from their outputs and trajectories. Continue with follow-up prompts until either model exceeds 1M cumulative tokens.
Read the evaluation workflowt
Completed engagementToptal
Computer-use evaluation & interaction data
Explored computer environments and recorded interaction data for LLM training and evaluation. Reviewed interaction patterns and data quality with the team.
Read the computer-use workflowALIGNERR / MODEL EVALUATION
Prompt. Compare. Follow through.
Design an open-ended engineering prompt, then compare two models’ outputs and execution trajectories. Write rubrics and evidence-based rationales, and develop follow-up prompts that test whether each model can sustain progress over an extended session.
Read the Alignerr workflow
High-level summaries of engineering responsibilities across these engagements.
03 / LEARNING & CONTENT CREATION
Complex ideas. Drawn simply.
I turn engineering concepts into visual lessons. Explore my growing collection on generative AI, agents and system design.
85 learning modulesExplore the library
30 MODULESGenAIStart with AI fundamentals. Explore generative models, prompting, embeddings, RAG and practical applications.Explore series 
25 MODULESAgentic AIExplore agent loops, planning, tools, memory and the design of systems that work toward a goal.Explore series 
30 MODULESSystem DesignTurn AI models into useful products. Define success, choose an architecture and connect data preparation with serving.Explore series 04 / BEHIND THE WORK
Build it.
Understand it.
Make it useful.
I’m Uday Kiran Reddy Kondreddy, an AI engineer based in Hyderabad. My work spans production retrieval systems, agent workflows, Python services, and coding-model evaluation.
I care about the part after the prototype: what breaks, how we notice, and whether our tests measure the behavior we actually need.
These project notes cover the problems I worked on, my engineering contributions and the systems behind them.
Have an interesting engineering problem? Let’s talkGOOD QUESTIONS WELCOME
Let’s compare notes.
Working through a retrieval problem, an agent workflow, or an evaluation that doesn’t quite add up?