Production agents
I build tool-using, conversational, RAG, and multi-agent systems that connect models to real business workflows.
PythonTypeScriptMCPRAGStrands
Atharv Kulkarni
India
Agentic AI Engineer at Onit
I build reliable production AI systems.
I turn ambiguous workflows into measurable agents using evaluation, observability, and disciplined software engineering.
Open to Applied AI and Agentic AI roles in the UK and globally.
System
Production AI loop
Live
- 01
Build
Agents · MCP · RAG
- 02
Evaluate
Golden sets · LLM judges
- 03
Observe
Traces · Scores · Feedback
- 04
Improve
Regression gates · Release
Shipping at enterprise scale
What I do
The model is only part of the system.
My work sits where AI behavior meets software engineering. I build the agent, define what good looks like, trace how it behaves in production, and use that evidence to make the next release better.
93%
agent accuracy
82%
lower processing time
56%
lower token usage
600+
enterprise customers
Evaluation and reliability
I turn product expectations into golden datasets, LLM judges, regression checks, and release decisions.
LLM-as-a-JudgeLangfuseGolden setsTracing
AI application engineering
I ship the software around the model: APIs, streaming, structured outputs, provider abstractions, and cloud runtimes.
FastAPISSELiteLLMAWS Bedrock
Selected work
Systems that moved a real metric.
A few examples of production AI, evaluation, and engineering automation from my recent work.
01
Onit · Production agent and evaluation system
Invoice Review Agent
93% accuracy82% faster56% fewer tokens
Rebuilt an agent that detects billing disputes, identifies disputed line items, and recommends invoice adjustments. I paired the new architecture with expert-labeled golden datasets and LLM-judge evaluation so performance could be measured before release.
Agentic AILLM evaluationGolden datasetsStrands
02
Onit · Legal operations analytics agent
Ask Unity
MCP data toolsProduction observabilityMulti-provider runtime
Built a conversational analytics agent that uses MCP data tools to answer spend and operational questions over enterprise data. Added multi-provider model support, production tracing, prompt and model comparison, and deployment on Amazon Bedrock AgentCore Runtime.
MCPLangfuseLiteLLMBedrock AgentCore
03
Onit · Multi-agent workflow
Requirements to user stories
~2 weeks to ~10 minutesStructured, reviewable output
Built a multi-agent system that converts long requirements documents into structured user stories, automating a high-effort product workflow while keeping the output reviewable.
Multi-agent systemsStructured outputsWorkflow automation
04
Onit · AI-assisted engineering
QA automation migration
~1 year of work in 3 weeksRecurring licensing cost removed
Drove the migration of QA automation across two enterprise product lines, using agentic coding for translation, refactoring, and validation while removing recurring tool licensing costs.
PlaywrightAgentic codingTest automation
Independent Projects
Product systems I built from scratch.
Independent work exploring developer infrastructure, reliable browser agents, and bounded multi-agent production workflows.
View all projects01Deployed prototype
AI developer platform
Skyeline
A platform for versioning prompts, routing LLM requests across providers, and observing the latency and token usage of agent and RAG applications.
Next.jsNestJSPostgreSQLDrizzle ORM
Read case study02Deployed prototype
Human-controlled browser agent
Relay
A desktop agent that fills browser forms while keeping review and final submission under explicit human control.
ElectronReactViteNestJS
Read case study03Advanced internal prototype
Agent-assisted publishing system
AI Guides Factory
An internal system for turning a guide idea into structured content, branded PDFs, privacy-safe screenshots, product listings, and marketing assets.
BunTypeScriptSQLitePlaywright
Read case studyExperience
Software foundations. AI focus.
Three years of hands-on GenAI work, backed by enterprise application engineering.
Onit
Agentic AI Engineer
June 2025 to present
Building and evaluating production AI agents for enterprise legal operations, with a focus on reliability, observability, and measurable releases.
Incubyte
Software Craftsperson, AI
October 2024 to June 2025
Developed enterprise AI products spanning multi-model chat, agents, RAG knowledge bases, deep research, streaming APIs, and document integrations.
IBM India
Application Developer
November 2022 to September 2024
Built enterprise software for testing, Workday integration monitoring, and data security, while exploring GenAI through projects and prototypes.
B.E. Computer Engineering
Honors in IoT · CGPA 9.31/10
International Institute of Information Technology, Pune · 2018 to 2022
Writing
Notes from the work.
2 min read
Building Agentic AI Systems
Lessons from building autonomous AI agents that can reason, plan, and execute multi-step tasks in production environments.
AI AgentsEngineering
2 min read
From Blockchain Research to LLMs
How publishing an IEEE paper on blockchain-based health records shaped my approach to building with large language models.
CryptoAI Agents