Sr AI Agent Engineer

Lexsi Labs; Bengaluru; Hybride; INR 1 200 000 - 1 800 000; Plein temps

Résumé du poste

LexsiLabs is building a high-autonomy AI Engineering Agent to execute end-to-end AI workflows atop Lexsi backend. The Sr AI Agent Engineer will define the core agent architecture with ownership from day one, deciding how the agent reasons, plans, and executes actions while ensuring interpretability and inspectability.

The role requires hands-on engineering, deep experience with agentic AI systems, and the ability to operate in ambiguous environments with a strong ownership mindset.

Qualifications

• Significant experience building and shipping complex AI/ML systems in production.
• Hands-on experience with agentic AI systems and frameworks (e.g., ReAct-, AutoGPT-like, LangChain, LangGraph, Semantic Kernel).
• Strong backend fundamentals: Python, APIs, databases, data pipelines, and cloud infrastructure.
• Systems-thinking mindset with consideration for performance, reliability, cost, safety, and interpretability.
• Comfort working in ambiguous, fast-moving environments with ownership and accountability.
• Bonus signals for long-horizon/stateful agents and enterprise/regulatory exposure.

Responsabilités

• Own the architecture and implementation of the AI Engineering Agent, shaping planners, memory, and execution control.
• Build robust tool orchestration and execution layers to interact with internal services and external APIs.
• Embed alignment, safety, and interpretability into system design through collaboration with the research team.
• Stress-test agent behavior in real-world conditions, identify edge cases, and iterate for reliability.
• Set a high engineering bar and contribute to design discussions, code reviews, and technical decisions.

Connaissances

AI/ML systems

Agentic AI frameworks

Python

APIs & services

Databases & pipelines

Cloud infrastructure

Systems thinking

Adaptability

Description du poste

LexsiLabsisafrontierAIlabfocusedonbuilding aligned,interpretable,andsafesuperintelligentsystems .Ourworkspansalignmentmethodologies,interpretability-ledsystemdesign,andfoundationalmodelresearchacrossstructured,tabular,andenterprisedata.WebuildAIsystemsintendedforreal-worlddeployment,wheretransparency,auditability,androbustnessarefirst-classconstraints.

Weoperatewithaflatstructure,highautonomy,andastrongbiastowardengineerswhotakefullownershipofwhattheybuild,fromarchitecturetoproductionbehavior.

TheRole

Wearebuildinga high-autonomyAIEngineeringAgent thatcanexecuteend-to-endAIengineeringworkflowsontopoftheLexsibackend.Thesystemisdesignedtoreasonovercomplexobjectives,planmulti-stepactions,interactwithtoolsanddatasystems,evaluateoutcomes,anditerateoverlonghorizonswhileremaininginterpretableandaligned.

Asa SrAIAgentEngineer ,youwillberesponsiblefordefiningandbuildingthecoreagentarchitecture.Youwillmakefoundationaldecisionsabouthowtheagentreasons,plansandexecutesactions,maintainsmemoryandstate,andhasitsdecisionsinspectedandexplained.Thisrolerequiresdeephands-onengineeringandtheabilitytooperateinambiguousenvironmentswithownershipfromdayone.

WhatYou’llBuild

• End-to-endAIEngineeringAgents thatcanautonomouslyperformtaskssuchasexperimentdesign,evaluation,analysis,andreporting,ratherthanmerelyassistingahumanintheseworkflows.
• HarnessEngineering: Abletodesignandmaintaincomplexharnesscomponentsthatsupportagents'scaleexponentiallyforcomplextasks.
• Agentarchitecturesthatcombinereasoning,planning,toolorchestration,andmemory ,withexplicithandlingoflong-horizonexecution,partialfailures,andself-correction.
• DeepintegrationswiththeLexsibackend ,includinginternalevaluationsystems,datapipelines,andalignmenttooling,aswellasenterpriseAPIsandproprietarydatasources.
• Agentsdesignedforenterpriseandregulatedenvironments ,whereeveryaction,decision,andoutputmustbeinspectable,explainable,anddefensible.

Responsibilities

• OwnthearchitectureandimplementationoftheAIEngineeringAgent ,makingearlydesigndecisionsaroundplanners,reasoningloops,memorymodels,andexecutioncontrolthatwillshapethesystemlong-term.
• Buildrobusttoolorchestrationandexecutionlayers ,ensuringtheagentcaninteractreliablywithinternalservices,externalAPIs,anddatasystems,andrecovergracefullyfromfailures.
• Embedalignment,safety,andinterpretabilityintothesystemdesign ,workingcloselywithresearchteamstoensurethatagentbehaviorisauditableandcontrollablebyconstruction.
• Stress-testagentbehaviorinreal-worldconditions ,identifyingedgecases,failuremodes,anddistributionshifts,anditeratingonthesystemtoimprovereliabilityandcorrectness.
• Setahighengineeringbar ,contributingtodesigndiscussions,codereviews,andtechnicaldecision-makingwithastrongsenseofownershipandaccountability.

Requirements

• SignificantexperiencebuildingandshippingcomplexAIorML-heavysystems ,whereyouhaveownedarchitectureandseensystemsevolveinproduction,notjustprototypes.
• Hands-onexperiencewithagenticAIsystemsandframeworks ,suchasReAct-styleagents,AutoGPT-likesystems,LangChain,LangGraph,SemanticKernel,orsimilar,withaclearunderstandingoftheirlimitationsandfailuremodes.
• Strongbackendengineeringfundamentals ,includingadvancedPythonproficiency,experiencebuildingAPIsandservices,andfamiliaritywithdatabases,datapipelines,andcloudinfrastructure.
• Asystems-thinkingmindset ,withtheabilitytoreasonaboutperformance,reliability,cost,safety,andinterpretabilityasinterconnecteddesignconstraints.
• Comfortworkinginambiguous,fast-movingenvironments ,whereproblemsarelooselyspecifiedandownershipisexpectedratherthanassigned.

StrongBonusSignals

• Experiencebuilding long-horizonorstatefulagentsystems thatmustreasonovertimeratherthansingle-turninteractions.
• Priorexposureto AIalignment,interpretability,orsafetytooling ,especiallyinproductionorenterprisesettings.
• Experienceworkingin regulatedorhigh-stakesdomains ,wheresystembehaviormustbeexplainedandauditedafterdeployment.
• Atrackrecordofdebuggingandimprovingsystemsthatbehavedunpredictablyintherealworld.

Wemovequicklyandexpectcandidatestodothesame.Wevaluesubstanceoverpolishandexecutionoverrhetoric.