Research Fellowship: Agent Intelligence & Evaluation

Ixigo; Gurugram District; Vor Ort; INR 2.500.000 - 6.000.000; Vollzeit

Zusammenfassung

Ixigo is seeking a research-focused professional to advance voice and dialogue systems. You will design evaluation frameworks, craft adversarial datasets across accents, and contribute to end-to-end observability of conversational interactions.

Ideal candidates hold a PhD in ML/NLP or related fields, are proficient in Python, and know speech models such as Whisper or Conformer, plus observability tools like OpenTelemetry or Langfuse.

Qualifikationen

• PhD candidates or equivalent track record in ML/NLP/speech research.
• Strong programming in Python and research-driven mindset.
• Familiarity with speech models, LLM tool-use, or observability stacks.

Aufgaben

• Design evaluation frameworks for voice and dialogue systems.
• Build datasets across accents and edge cases for robust testing.
• Develop end-to-end observability across audio, STT, LLM, and TTS paths.

Kenntnisse

Python

ML/NLP intuition

Speech systems familiarity

Research mindset

Ausbildung

PhD in ML/NLP/Speech

Tools

Whisper

Conformer

OpenTelemetry

Langfuse

Arize

Hamming

Jobbeschreibung

Job Description

Voice agents fail in ways traditional software doesnt. An ASR confidence drop on a regional accent misfires a tool call, an LLM hallucinates a policy because upstream latency broke turn-taking, and support teams roll these agents back within a week without anyone able to explain what went wrong.

What youll work on

Over 4 months, youll take on one or two of the following, shaped by your interests.

Evaluation frameworks.

Text-only evals miss most of what matters in voice: barge-in, prosody, latency-induced errors, cross-turn context loss. Youll design audio-native metrics, generate adversarial conversational datasets across accents and edge cases, and build LLM-as-judge rubrics for task completion, empathy, and recovery from tool failures.

End-to-end observability.

Tracing a failed interaction means correlating audio packets, STT hypotheses, LLM reasoning traces, tool calls, and TTS output back to a single conversation ID. Youll help shape the schema and analysis layer that makes cascade failures visible across the stack.

Self-improvement systems.

Once you can measure and trace, the interesting work is closing the loop: mining production traces for failure patterns, generating targeted fine-tuning data or prompt updates, and validating that fixes hold under adversarial replay.

Qualifications

Who were looking for

Someone who cares about the research questions for their own sake, and equally cares whether the work ships. Papers at Interspeech, ACL, NeurIPS, or EMNLP on speech, dialogue systems, agent evaluation, or human-AI interaction are directly relevant.

Comfortable in Python, and familiar with at least one of: speech models (Whisper, Conformer variants), LLM tool-use and agent frameworks, or observability stacks (OpenTelemetry, Langfuse, Arize, Hamming). Current PhD students in ML, NLP, or speech are the strong default; exceptional MS students or research engineers with a publication track record are welcome to apply.

Nice to have

Prior work on evaluation methodology, dataset synthesis, or interpretability. Experience with real-time systems, telephony, or streaming pipelines. A blog, repo, or workshop paper that shows how you think in public.

Disclaimer: This job posting has been aggregated from external source. Role details, content, and availability are subject to change. Applicants are advised to confirm the latest information directly on the company website before applying.