Research Scientist - New Delhi - Zigment

Description

Why Zigment.ai
In today's fast-paced digital landscape, engaging and converting leads is crucial for every business. Zigment.ai provides next-generation AI conversational solutions to help businesses grow their revenue, acquire more customers, and build a stronger brand with ease.

We see a massive opportunity ahead and are committed to building a significant, lasting business. We're assembling a diverse team of curious, creative individuals who find purpose in their work and support one another. If this resonates with you, we'd love to hear from you.

Read more on

https://www.zigment.ai

About the Team

The AI Research team at Zigment.ai is responsible for the core intelligence behind our conversational AI platform. We research, design, and build the ML and LLM systems that power our AI agents — from language understanding and generation to reasoning, personalisation, and domain adaptation. Our work directly shapes the product and pushes what's possible in AI-driven business conversations.

About the Role

As a

Research Scientist / Research Engineer

at Zigment.ai, you will advance the capabilities of our AI systems through original research and hands-on model development. Your work will span the full research lifecycle — formulating hypotheses, designing experiments, training and fine-tuning models, analysing results, and translating breakthroughs into production-ready systems.

We're looking for people who have strong foundations in machine learning and NLP research, can iterate quickly on ideas, are proficient at both writing research code and engineering robust systems, and are excited about building AI that solves real business problems at scale.

This is a WFO Role in Bengaluru, KA.

You might thrive in this role if:

You love being on the cutting edge of LLM, NLP, and machine learning research.
You're a self-starter who takes ownership of research directions and drives them from idea to implementation.
You value principled approaches — simple experiments in tightly controlled settings that lead to trustworthy, reproducible conclusions.
You thrive in a fast-paced, dynamic environment where rapid iteration and creative problem-solving are key.
You're comfortable diving deep into large ML codebases to debug, improve, and extend them.
You have a deep understanding of how modern language models work and a strong intuition for what makes them better.

What You'll Need

Master's or PhD in Computer Science, Machine Learning, AI, or a closely related field (or equivalent practical experience with a strong research track record).
Deep understanding of modern LLM architectures — Transformers, attention mechanisms, decoder-only models, mixture-of-experts, etc.
Hands-on experience training, fine-tuning, or adapting large language models (techniques such as LoRA, QLoRA, RLHF, DPO, SFT, or similar).
Strong proficiency in Python and deep learning frameworks (PyTorch preferred; JAX a plus).
Experience designing and executing rigorous ML experiments with proper evaluation, ablation studies, and reproducibility.
Familiarity with the Hugging Face ecosystem (Transformers, PEFT, Datasets, TRL) or equivalent tooling.
Ability to read, critically assess, and implement ideas from recent research papers.
Strong mathematical foundations in probability, statistics, optimisation, and linear algebra.

Preferred Qualifications

Publications or significant contributions at top-tier ML/NLP venues (NeurIPS, ICML, ICLR, ACL, EMNLP, etc.).
Experience with reinforcement learning from human feedback (RLHF), reward modelling, or alignment research.
Familiarity with distributed training and scaling strategies (DeepSpeed, FSDP, Megatron-LM).
Experience building or improving inference pipelines (vLLM, TensorRT-LLM, ONNX Runtime).
Knowledge of retrieval-augmented generation (RAG), agentic frameworks, or tool-use in LLMs.
Experience with multi-modal models (vision-language, speech-language).
Contributions to open-source ML/AI projects or publicly shared research artefacts.
Prior experience in a startup or fast-paced product environment where research directly impacts production.

What You'll Do

Research & Experiment

— Formulate research questions, design experiments, and develop novel techniques to improve the quality, reliability, and reasoning capabilities of our LLM-powered systems.
Train & Fine-Tune Models

— Build and run training pipelines for fine-tuning foundation models on domain-specific tasks such as conversational sales, lead qualification, intent understanding, and personalised engagement.
Develop New Architectures & Methods

— Explore and prototype new model architectures, training objectives, decoding strategies, and data curation techniques that push our AI forward.
Evaluate Rigorously

— Design comprehensive evaluation frameworks including automated benchmarks, human evaluation protocols, and production A/B tests to measure model performance and guide research decisions.
Build Research Infrastructure

— Create and maintain tools, datasets, and pipelines that accelerate experimentation and enable the team to move faster.
Publish & Share Knowledge

— Document findings, write internal research reports, and, where appropriate, contribute to the broader ML community through publications and open-source work.
Collaborate Across Teams

— Work closely with engineering and product teams to ensure research insights translate into real product improvements deployed at scale.
Stay Current

— Continuously engage with the latest research in LLMs, NLP, RL, and related fields; bring new ideas to the team and evaluate their applicability.

Critical Soft Skills

Intellectual curiosity

— a genuine passion for understanding how models learn, fail, and improve, and a drive to push the state of the art.
Scientific rigour

— commitment to clean experiments, reproducible results, and honest interpretation of findings.
Clear communication

— ability to distil complex research into clear narratives for both technical and non-technical audiences.
Ownership & initiative

— comfort with ambiguity and the ability to independently identify high-impact research directions and see them through.
Pragmatism

— knowing when to pursue research depth vs. when to deliver a working solution; balancing exploration with business impact.
Collaboration

— ability to work effectively in a multidisciplinary team, give and receive constructive feedback, and build on others' work.

If you don't think you meet all the criteria above but are still interested in the opportunity, please apply It's impossible for one person to check every box, and we're looking for someone who is willing to grow with us and create impactful work

Ready to Apply?

Zigment.ai has a positive, diverse, and supportive culture — we look for people who are curious, inventive, and work to be a little better every single day. In our work together, we aim to be smart, humble, hardworking, and, above all, collaborative. If this sounds like a good fit for you, why not say hello?