Software Engineer - 2 (Voicebot)

Exotel Techcom Pvt Ltd; Bengaluru; On-site; INR 1,500,000 - 2,500,000; Full time

Job summary

Exotel Techcom Pvt Ltd is hiring for an engineering role focused on building and maintaining AI-powered voicebot capabilities. You will evaluate models, perform fine-tuning with PEFT/LoRA/QLoRA on open-weight models, and benchmark across providers and self-hosted setups.

You will own the end-to-end SDLC, from design through deployment and monitoring, ensuring low latency and high-quality conversational experiences in production.

Qualifications

• Solid grounding in ANN and transformer architectures.
• Experience with LLM eval harnesses, regression suites.
• Fine-tuning with PEFT/LoRA/QLoRA on open weight models.
• Speech/ASR-TTS evaluation: WER, latency, diarization.
• Strong Python and production-grade coding skills.

Responsibilities

• Build and maintain LLM/speech eval frameworks for the voicebot.
• Run fine-tuning experiments on open-weight models.
• Benchmark LLMs and ASR/TTS across providers and self-hosted options.
• Diagnose production conversation failures using logs and traces.
• Own SDLC changes from design to monitoring.

Skills

Python

LLM evals

Deep Learning

NLP

Tools

PEFT/LoRA/QLoRA

Job description

AboutUs

ExotelisaleadingproviderofAItransformationtoenterprisesforcustomerengagementandexperience.Withover20billionannualconversationsacrossOmnichannel,voice,agents,andbots,Exotelistrustedbymorethan7000clientsworldwide,spanningindustriessuchasBFSI,Logistics,ConsumerDurables,E-commerce,Healthcare,andEducation.

Customerexpectationsareevolving,andbusinessesfacethechallengeofbalancingtheneedforincreasedrevenue,optimizedcosts,andexceptionalcustomerexperience(CX).Exotelstepsforwardasyourtransformativepartner,offeringanAI-poweredcommunicationsolutiontoaddressallthree!

AI@ExotelVoicebot

TheVoicebotteambuildsandoperatesExotel'sreal-timevoiceAIproduct—productionbotshandlinglivephoneconversationsforenterprisecustomers.

We run real - time conversational pipelines end to end: speech recognition → LLM reasoning/orchestration → speech synthesis, with tool-calling for backend actions.

We evaluate and swap models constantly across providers, on cost, latency, and conversation quality not vibes.

Webelieveinmeasuringwhatmatters:agooddemoisn'tthesameasagoodeval.

TheRole

You'llbepartoftheteambuildingandcontinuouslyimprovingExotel'svoicebot—fromthemodellayer(fine-tuning,evals)totheliveconversationexperience(speechquality,latency,turn-taking).Thisisanengineeringrolefirst:you'llbuild,evaluate,andshipchangesthatdirectlyimprovecallqualityandbusinessmetricsinlivecustomerdeployments.

WhatWeExpectatThisLevel

Independentexecution.Givenascopedproblemandanagreedapproach,youtakeittoproductiononyourown—build,eval,deploy,monitor—withoutneedingtobeunblockeddaily.

Deepownershipofevalframeworksandworkingknowledgeofassociatedservices/infra.YougodeepontheAIsideoftheproduct—models,evals,speechquality—andknowenoughaboutthesurroundingservicestotracealiveproblemacrossthepipelineandseeitthrough.

WhatYou'llDo

Build and maintain LLM/speech eval frameworks for the voicebot — task success, hallucination, instruction-following, WER/latency, barge-in and turn-taking quality — across model and prompt changes.

Run fine-tuning experiments (full FT, PEFT/LoRA/QLoRA) on open-weight models for domain specific voicebot tasks, and produce the evidence for when fine-tuning beats prompting.

Benchmark LLMs and ASR/TTS engines on cost, latency, and quality across providers and self hosted options and make a clear recommendation from the data.

Diagnose and fix real production conversation failures bad turn taking, misrecognition, latency spikes, prompt regressions using logs, traces, and eval data, not guesswork.

Shipchangesintotheliveconversationalpipeline,withinstrumentationandalertingbuiltinfromdayone.

Take ownership across the SDLC for your changes: design (with a senior engineer), eval design, deployment, and monitoring.

WhatYouBring

Must-have

Solid grounding in ANNs and transformer architecture attention, tokenization, decoding strategies enough to reason about why a model behaves a certain way, not just call an API.

Hands-on experience with LLM evals: building or running eval harnesses, LLM-as judge setups, regression suites for prompt/model changes.

Hands-on experience with fine-tuning, including PEFT/LoRA/QLoRA — on at least one open weight model, for a real task (not just a tutorial).

Working knowledge of speech/ASR - TTS evaluation — WER, latency, diarization, common failure modes in real (noisy, accented, multilingual) audio.

StrongPython;comfortablereading/writingproductioncode,notjustnotebooks.

2-4 years of software/ML engineering experience, with at least some of it in a production system (not purely research/academic).

A track record of shipping and owning your own changes in production you've been on the hook for something live.

Stronganalyticalrigor—youinstinctivelyask"howdowemeasurethis"beforeshippingachange.

Good-to-have

Experiencewithreal-timeaudio/streamingsystemsandstreamingvs.batchtradeoffs.

Experiencewithagenticorchestrationandtool-callingpatternsforLLMs.

ExposuretoRAGpatterns—embeddings,vectorstores,retrievalstrategies.

Familiaritywithself-hosting/servingopen-weightmodels.

FamiliaritywithobservabilityforAIworkloads—costtracking,qualitydashboards.

Experiencewithmulti-tenantSaaSconstraints(per-tenantconfig,isolation).

PriorexperiencespecificallyinvoiceAI/IVR/contact-centerdomains.

HowWeWork

Youownit.Build,eval,ship,monitor—andstayonthehookwhenit'srunninglive.There'snoseparate"MLOpsteam"tohandoffto.

Youmeasureit.Nomodelorpromptchangeshipswithoutanevalstorybehindit.

Youcollaborate.You'llworkcloselywiththevoicebotarchitecture/productteamandfielddeliveryengineersshippingtoenterprisecustomers.Goodideaswinregardlessofsource.

Youstaycurious.Newmodelsandtechniqueslandconstantly—evaluatingandbenchmarkingnewoptionsispartofthejob,notasideproject.

WhyExotel

Work on AI problems at real scale: live voice conversations, not offline batch jobs, for enterprise customers.

Strongseniorengineerstodesignwith,andrealownershipofwhatyoubuild.

Ateamthattreats"doesitactuallywork"asmoreimportantthan"doesitdemowell".

Opportunity to work across the full voicebot AI stack: LLMs, speech, real time orchestration, and the infra it runs on.