About the Programme

Developing an enterprise AI platform focused on financial compliance and intelligence

Role Overview

We are looking for an LLM / AI Engineer with 5 years experience to build the language model layer of the platform. You will design and implement the systems that enable the platform to understand compliance documents, retrieve relevant legal and regulatory knowledge, classify and triage notices, and generate structured outputs such as response drafts and recommendations. Accuracy and reliability are non-negotiable in a compliance context — you will be responsible for ensuring that every LLM output is grounded, explainable, and validated before it reaches the compliance team.

Key Responsibilities

Agentic AI abd RAG Architecture

- Design and build the AI orchestration layer that coordinates multi-step compliance workflows
- Build LLM-based document intelligence modules that extract, validate, and cross-reference structured data from financial documents to identify discrepancies and produce actionable fault reports.
- Design and build the RAG (Retrieval-Augmented Generation) architecture that powers the platform’s legal and regulatory knowledge layer.
- Build the knowledge base ingestion pipeline — parsing, chunking, embedding, and indexing legislation, circulars, tribunal rulings, and past compliance responses.
- Design the retrieval layer to surface the most relevant knowledge for each query — including re-ranking, hybrid search, and context window management.
- Maintain and expand the knowledge base as regulations evolve, ensuring the system reflects current law at all times.

Document Understanding & Classification

- Build document understanding capabilities for compliance documents — invoices, notices,



refund applications, and government portal outputs.
- Design and implement classification systems for incoming notices — identifying issue type, relevant period, amount at stake, and statutory deadline from unstructured document content.

Structured Output Generation

- Design prompt engineering frameworks that produce structured, reliable outputs suitable for use in a compliance workflow
- Build response drafting capabilities for common notice types, grounded in the knowledge base and producing outputs with clear citation trails.
- Ensure all generative outputs include the source references used — the compliance team must be able to verify every recommendation the system makes.

Evaluation & Quality Assurance

- Build a systematic evaluation framework to measure LLM output quality — accuracy, groundedness, citation correctness, and consistency across similar inputs.
- Identify and mitigate hallucination risks — implement guardrails, confidence scoring, and fallback behaviours for low-confidence outputs.
- Continuously monitor output quality in production and feed findings back into prompt and retrieval improvements.

Required Qualifications

Education

- B.E. / B.Tech / M.Tech in Computer Science, Information Technology, or a related field.

Experience





- 5+ years of engineering experience with at least 3 years focused on LLM systems, NLP, or applied AI in a production environment.
- Proven experience building and deploying RAG systems in production — not research prototypes.
- Experience working with legal, regulatory, or technical document corpora is strongly preferred.

Technical Skills

- LLM Frameworks: LangChain, LlamaIndex, or equivalent for building RAG and agentic pipelines.
- Vector Databases: Databricks Vector Search, Pinecone, Weaviate, or equivalent; experience with hybrid search approaches.
- Embedding Models: Familiarity with a range of embedding models and trade-offs between them for domain-specific retrieval.
- Prompt Engineering: Structured output generation, chain-of-thought prompting, few-shot design, and systematic prompt evaluation.
- Evaluation: LLM evaluation frameworks — RAGAS, TruLens, or equivalent; experience designing domain-specific evaluation datasets.
- Document Processing: PDF parsing, OCR, and document chunking strategies for varied document types and layouts.
- Languages: Python proficiency; familiarity with the Hugging Face ecosystem.

Preferred Qualifications

- Prior experience in legal technology, compliance, tax, or regulatory document processing.
- Familiarity with Indian regulatory document formats — GSTN notices, CBIC circulars, or tribunal orders.
- Experience with fine-tuning or domain adaptation of language models for specialised corpora.
- Knowledge of agentic AI patterns for multi-step compliance workflows.
- Experience building multilingual document understanding systems.

Pay: ₹1,000,000.00 - ₹1,500,000.00 per year

Advantages

- Provident Fund

Work Location: In person

📌 LLM / AI Engineer (Thane)
🏢 Quantorus 365
📍 Thane