Machine Learning Engineer

About Aiverbalyze

Aiverbalyze Technologies is a deep-tech communication intelligence company building domain-adapted AI models that humanize machine conversations. Our proprietary stack includes optimized LLMs, multilingual ASR/TTS, Visual Language Models, and Neural Machine Translation all trained on region-specific, domain-rich datasets covering 30+ Indian and global languages.

We power intelligent voice, chat, and email automation for enterprises across BPO, healthcare, banking, and telecom — running on both edge and cloud with quantized deployments for high-throughput, ultra-low-latency inference.

Role overview

We are looking for a passionate and technically sharp ML Engineer Intern to join our core AI team as paid internship with full time job prospects. You will work directly on training, fine-tuning, and optimizing large language models, speech models, and multimodal architectures — and deploying them at production scale. This is a hands-on, deep-tech role with real ownership from day one.

Outstanding interns will receive a full-time job offer upon completion.

What you will work on

• Fine-tuning and domain-adapting LLMs (Qwen, LLaMA, Mistral variants) for enterprise conversation use cases
• Training and optimizing multilingual ASR/TTS models using NeMo, ESPnet, or similar frameworks
• Running distributed training jobs (DDP/FSDP) on multi-GPU clusters and debugging training instabilities
• Implementing quantization (INT4/INT8 via AWQ, GPTQ, BitsAndBytes) and running benchmarks for accuracy-latency tradeoffs
• Optimizing inference pipelines using vLLM, TensorRT-LLM, or ONNX Runtime for low-latency deployment
• Building and curating domain-specific datasets — cleaning, deduplication, augmentation, and formatting for instruction tuning
• Evaluating models on real-world benchmarks: WER, BLEU, ROUGE, and custom domain evals
• Collaborating with the product and platform team to deploy models via Dockerized services and REST APIs

What we are looking for

• Final year B.Tech / M.Tech / MS student in CS, EE, Data Science, or a related field — or a recent graduate
• Strong Python skills and comfort with PyTorch; experience with HuggingFace Transformers is a must
• Understanding of transformer architectures (attention, positional encoding, tokenization)
• Familiarity with training concepts: loss functions, optimizers, learning rate schedules, gradient checkpointing
• Exposure to at least one of: NeMo, vLLM, TGI, TensorRT, ONNX — even from personal projects
• Comfortable working in Linux/Docker environments and reading model/system logs
• Curiosity, ownership mindset, and ability to read research papers and translate them to code

Good to have

• Experience with speech models — Whisper, NeMo FastConformer, Wav2Vec, or similar
• Knowledge of RLHF, DPO, or instruction tuning pipelines
• Familiarity with multilingual NLP — Indian language datasets, transliteration, code-switching
• Prior internship, research project, or Kaggle experience in NLP or speech
• Contributions to open-source ML repos