Skip to content
Search jobs inside your AI agent. Claude, ChatGPT, Claude Code, Codex, Gemini, Cursor and Grok
Connect

Convin

NewBe an early applicant

Machine Learning Engineer

Posted

Summary

Convin (Bengaluru) seeks a Machine Learning Engineer (2–5 yrs) to own the full ML lifecycle: fine-tuning small language models for call analysis, building LLM inference/serving infrastructure (NVIDIA Triton, vLLM, SGLang), and improving multilingual ASR while developing an in-house TTS model. Core stack: Python, PyTorch, Hugging Face, GPU inference optimization.

Job Title: Machine Learning Engineer

Experience: 2–5 Years

Location: Bengaluru

Role Type: Full-time | Individual Contributor


About Convin:

  • Convin is building AI that turns customer conversations into measurable business outcomes. Founded by IIT Delhi alumni, Convin develops proprietary AI products that help B2C businesses engage customers across phone calls, WhatsApp, SMS, and website voice widgets. Convin’s AI products understand customer intent and determine the right channel, timing, message, and next action for every interaction.
  • Convin helps businesses turn customer interest into qualified leads, booked appointments, renewals, and completed payments. Convin is building for two demanding B2C markets, India and the United States. In India, we serve leading EdTech, retail, and BFSI businesses. In the United States, we help home service businesses turn more inquiries into booked jobs.
  • At Convin, every team contributes to solving real business problems. Whether you are building our products, taking them to market, growing customer relationships, or strengthening the organization behind them, your work will directly influence Convin’s growth and customer outcomes.
  • As Convin scales, you will have the opportunity to take ownership, work across India and US markets, and help shape the products, teams, and systems behind our growth. Convin is trusted by leading brands including Physics Wallah, SBI Life, Shaadi.com, Bajaj, Niva Bupa, Aditya Birla & RBL.


Highlights:

  • Raised $6.5 million in Series A funding in 2024
  • Ranked No. 6 on LinkedIn’s Top Startups 2025 list in India
  • Ranked the 19th Best Software Company in India by G2


We are looking for a Machine Learning Engineer with expertise across Speech and Language Models, who can take ownership of model development, fine-tuning, evaluation, optimization, and production deployment.

The role involves working on Small Language Models (SLMs), LLM inference infrastructure, multilingual Automatic Speech Recognition (ASR), and developing an in-house Text-to-Speech (TTS) model.


Key Responsibilities:


1. Small Language Models (SLMs) & LLM Fine-tuning

  • Fine-tune, distill, and maintain custom SLMs for call analysis, structured information extraction, and AI agent reasoning.
  • Build and curate training and evaluation datasets, including synthetic data generation and data-labeling pipelines.
  • Conduct systematic model evaluations using task-specific metrics and benchmark against larger reference models.
  • Optimize models for quality, latency, and cost using quantization, knowledge distillation, LoRA, and QLoRA.


2. LLM Inference & Serving Infrastructure

  • Develop, deploy, and optimize LLM serving infrastructure using NVIDIA Triton Inference Server, vLLM, SGLang, and NVIDIA Dynamo.
  • Optimize inference performance for low-latency, high-concurrency voice workloads.
  • Implement batching, prefix/KV caching, quantized weights, and speculative decoding.
  • Build model routing, health checks, fallback mechanisms, and production monitoring.
  • Troubleshoot GPU memory, throughput, latency, and model reliability issues.


3. Speech Technologies – ASR & TTS

  • Evaluate, fine-tune, and improve multilingual ASR systems, particularly for Indian languages and code-mixed speech.
  • Research and benchmark open-source TTS architectures for naturalness, latency, streaming capabilities, and language coverage.
  • Lead the development and training of an in-house TTS model.
  • Work on speech data collection, cleaning, speaker characteristics, prosody, and model evaluation.
  • Evaluate speech models on Mean Opinion Score (MOS), intelligibility, and real-time inference performance.
  • Deploy and optimize speech models for production voice applications.


Required Skills & Qualifications:


Must-Have

  • 2–5 years of experience in Machine Learning, NLP, Deep Learning, or related domains.
  • Strong programming skills in Python and hands-on experience with PyTorch.
  • Experience training and fine-tuning Transformer-based models.
  • Practical experience with LLM fine-tuning and evaluation using Hugging Face, PEFT, or similar frameworks.
  • Experience deploying and maintaining ML models in production environments.
  • Hands-on production experience with at least one of NVIDIA Triton Inference Server, vLLM, or SGLang.
  • Understanding of inference optimization, GPU utilization, latency, throughput, and serving costs.
  • Strong problem-solving skills and experience working with real-world datasets.


Good to Have

  • Experience with ASR architectures such as Whisper, CTC, or RNN-T.
  • Experience with TTS architectures such as VITS, FastSpeech, flow-matching, or codec-based models.
  • Exposure to NVIDIA Dynamo or distributed inference frameworks.
  • Experience working with Indian languages and code-mixed speech.
  • Knowledge of quantization techniques such as FP8, AWQ, and GPTQ.
  • Experience with distributed training frameworks such as FSDP and DeepSpeed.
  • Familiarity with Docker, Kubernetes, Prometheus, and Grafana.
  • Open-source contributions or research publications in NLP, Speech, or Machine Learning.


What We Offer

  • Opportunity to work on cutting-edge Conversational AI and Generative AI technologies.
  • Ownership of end-to-end ML projects, from research and experimentation to production deployment.
  • Exposure to emerging Speech and Language Model architectures.
  • Opportunity to solve challenging problems in multilingual AI and real-time voice applications.


Skills

Apply

See also

ML / AI jobs by country — openings, pay and top skills →