Machine Learning Intern

• Conduct experiments and reports results reliably, with guidance

• Experiments includes (but not limited to):

• Benchmark open-source LLMs (gpt-oss-120/20b, Llama models, etc.) against the proprietary LLMs (like OpenAI's GPT-4o/-mini, Gemini, etc.)

• Experiment with LLMs to create the scalable and search-efficient KB, synthetic QA generation from the digital documents, and Prompting optimizations to build the scalable chatbots

• Evaluate language translation models/services for Indian languages (e.g., Bhashini, Sarvam, Google Translate, etc.)

• Assess Speech models for Indian languages for ASR (STT) and TTS tasks (e.g., Amazon Polly, AI4Bharat's Conformers, Sarvam, etc.)

• Improve the existing RAG-QA pipeline by finding performance gaps, benchmark different Retrieval (Embedding) and chunking techniques

• Gather, clean, analyze, and process the text and speech data for building the knowledge base (KB) for conversational chatbots.

• Learns to derive insights from experiments and next steps

• Watches incoming data regularly and performs quality checks

• Collaborates with cross-functional teams to complete tasks on time

• Proactively seeks help and required information from peers

• Communicate the research findings in a clean and compact manner

• Supports development of good and clean codebases with documentation of code and work consistently with high standards

• Communicates and presents results effectively with peers

• Stay updated with recent advancements in GenAI-LLMs, ASR (STT and TTS), RAG-QA, Evaluation of LLMs, etc. (that can be applied in our product)

• Develops expertise with typical ML tooling such as Pandas, ML frameworks (Pytorch, Scikit-Learn), Excel (Pivot tables), Visualization libraries, Experiment monitoring (Weights & Biases), GitHub

• Learns to work efficiently with tooling: Unix, VSCode, Google office suite, Calendar, Slack

• Ability to work in a fast-paced startup environment

• Eagerness to learn, and apply the latest research/work happening in the domain in the solution

• Good at AI/ML fundamentals

• Good at LLM fundamentals like RAG-QA, Prompt Engineering, Evaluations, Vector Stores, Retrievals and Chunking, Fine-tuing, Synthetic data generation, model deployment, etc.

• Experience with GenAI tools (but not limited to) like LangChain, LlamaIndex, LlamaParse, Langfuse, FAISS, Chroma, vLLM, OpenAI toolkits and SDK, etc.

• Familiarity with OpenAI models and tools, Open-source LLM models like gpt-oss, Llama, Gemma, Mistral, Wave2Vec2, Bhashini's or AI4Bharat's language translation model, etc.

• Familiarity with Docker, AWS, GCP is a plus

• Strong Python coding and debugging skills, hands-on experience with some of the Data Science toolkits like Pandas, Numpy, Matplotlib/Seaborn, etc. and preferably at least one Deep Learning Framework among Pytorch (preferably), Keras, TensorFlow, etc.

• Should have completed coursework in Probability, Linear Algebra, Calculus and preferably has some exposure to AI / Machine Learning.

• Highly preferred to have demonstrated experience of working in the field via an internship or project. Do provide links to some of your open-source projects.

• Prior exposure to Linux/Unix is expected before joining for the internship.

WadhwaniAI LEHS