I build open-source tooling for agentic AI systems — the infrastructure layer that makes AI agents more reliable, testable, and production-ready.
Currently based in Bengaluru 🇮🇳 | Open to AI/ML engineering roles
I focus on the gap between prototype agents and production agents — fault tolerance, search-based reasoning, and testing infrastructure for LLM-powered systems.
agentchaos-tools — Chaos testing for agentic AI
Fault injection hooks for openai-agents-python. Simulate tool failures, latency spikes, and corrupted outputs to verify your agent handles real-world conditions gracefully.
from agentchaos import FaultInjectionHooks, ToolFault, FaultType
hooks = FaultInjectionHooks(
faults=[
ToolFault(tool_name="web_search", fault_type=FaultType.EXCEPTION, rate=0.5),
ToolFault(tool_name="calculator", fault_type=FaultType.LATENCY, latency_seconds=2.0),
],
seed=42,
)
result = await Runner.run(agent, "Search for something", hooks=hooks)
hooks.report()
# agentchaos report — 1 fault(s) triggered:
# 1. [EXCEPTION] tool='web_search'pip install agentchaos-toolssmolagents-lats — Tree search reasoning for smolagents
Implementation of LATS (Language Agent Tree Search) for HuggingFace's smolagents. Replaces the linear ReAct loop with a UCT-guided search tree — agents expand branches, score them, and reflect on failed paths before backtracking.
from smolagents_lats import LATSAgent
agent = LATSAgent(tools=[], model=model, n_branches=3, max_depth=4)
result = agent.run("What is the 10th Fibonacci number?")pip install smolagents-lats
AI / Agents: openai-agents-python · smolagents · crewAI · strands-agents · LangGraph
Testing & Reliability: chaos engineering · fault injection · pytest
Concepts: ReAct · LATS · MCTS · multi-agent systems · RAG
I'm actively building in the agentic AI space and always open to interesting conversations, collaborations, or opportunities.