Job Summary

Position: AI Engineer. Education: Bachelors or Masters degree in Computer Science, Artificial Intelligence, Data Science, Information Technology, or a related discipline. Experience range: 3-7 Years. Employment type: Full-time. Location: Mumbai.

We are hiring an AI Engineer to join our AI Practice in Mumbai. This is an applied AI engineering role. You will spend roughly 60% of your time building and extending Scout, GoldenSource's AI platform - a production system with real users, running against real client data.

The remaining 40% is practice work: taking AI capability into client engagements, proving out new use cases, and turning what works into something the product can absorb. Those two halves ask for different things. Product work rewards depth, rigour and patience with an existing codebase.

Engagement work rewards range, speed, and comfort with unfamiliar ground. We are looking for someone who is genuinely interested in both, and honest about which one they find easier. You do not need a background in capital markets or financial data.

We will teach you the domain. What you cannot pick up on the job is the engineering judgement - knowing when retrieval is the answer and when it isnt, and how to get an AI application into production and keep it there.

The AI Practice is currently five people and growing. That means the work has real surface area and few layers between you and the decisions - but also that some of our engineering practices are still being established, and you will be expected to help shape them rather than inherit them. If you want a fully formed process to join, this is not yet that team. If you want to influence how a practice is built, it is.

Responsibilities

- Scout. Extending the platform's agentic capabilities - retrieval over enterprise data stores, tool integration via Model Context Protocol, orchestration across multi-step workflows, and the evaluation harnesses that tell us whether any of it actually works.
- Retrieval and grounding. Designing how the system finds, ranks and grounds answers across relational and analytical data sources, and improving that as the corpus and the question types grow.
- Agents and orchestration. Building agentic workflows that call tools, query data, and hand off between steps reliably enough to run unattended.
- Evaluation. Establishing how we measure quality, accuracy and regression for non deterministic systems.



This is a first-class part of the job, not an afterthought.
- Production engineering. APIs, services, deployment, monitoring, cost and latency. Work that ships, not work that demos.
- Client engagements. Working alongside Product Management and Professional Services to shape AI use cases for clients, prototype quickly, and identify what is worth productising.

Required Qualifications

- Bachelors or Masters degree in Computer Science, Artificial Intelligence, Data Science, Information Technology, or a related discipline.
- 3-7 years of professional software engineering experience, with meaningful recent time spent building LLM-based applications.
- At least one AI application you have taken to production and supported afterwards.
- Robust software engineering fundamentals - you write code others can maintain.

Required Technical Skills

- Core Python, to a professional standard.
- SQL, including working comfortably against large relational datasets.
- Large Language Models in practice: prompt design, context management, structured outputs, cost and latency trade-offs.
- Retrieval-Augmented Generation - chunking, embedding, vector search, re-ranking, and the failure modes of each.
- Agentic patterns: tool use, function calling, multi-step workflows.
- At least one agent or LLM framework in anger (LangChain, LangGraph, Llamalndex, or equivalent).
- Vector stores (pgvector, FAISS, Pinecone, Weaviate, Milvus, or similar).
- REST API development (FastAPI or Flask).
- Cloud platform experience - AWS, Azure or GCP.
- Docker, and familiarity with deploying containerised services.
- Git and modern CI/CD practice.
- Fine-tuning approaches - full fine-tuning, LoRA/PEFT, instruction tuning - and, more importantly, the judgement to argue when fine-tuning is the right answer against prompting, retrieval, or better data.
- Model evaluation for generative systems: benchmarks, LLM-as-judge, human review, regression testing.




- Foundation model landscape - open and commercial - and enough currency to have an informed opinion on what has changed in the last six months.
- Classical machine learning fundamentals sufficient to know when a simpler model beats an LLM.
- Data handling with Pandas and NumPy.
- Responsible AI: guardrails, evaluation for harm, explainability, and the governance expectations of regulated clients.

Nice to Have

- Knowledge of capital markets, securities reference data, or financial data management.
- Experience embedding AI into enterprise SaaS products.
- Hands-on fine-tuning of open-source models such as Llama, Mistral or Qwen.
- Model Context Protocol (MCP) - building servers, clients, or gateways.
- Workflow automation and orchestration tooling.
- Graph databases and knowledge graphs.
- Kubernetes, Mlflow, or comparable MLOps tooling.
- Experience working in Agile/Scrum delivery.

Who We're Looking For

- We are looking for someone who works at startup pace but thinks in product terms. In practice that means you think in whole solutions, you ask what it is for, who uses it, and what happens after it ships - not only how to build the piece in front of you.
- You are comfortable pushing back on a requirement when the underlying question is wrong.
- You own things end to end. You take a problem from ambiguity through to something running in production, and you stay with it afterwards. Nothing here is finished at the pull request.
- You are honest about what you have built. You say when something doesn't work, when a result is uncertain, and when you don't know. In a field where systems produceconfident-sounding output that is wrong, this matters more than usual.
- You hold a quality line under time pressure. Startup pace does not mean shipping whatever passes. We expect the speed and the standard together, and we expect you to say clearly when the two are genuinely in conflict rather than quietly trading one away.
- You communicate clearly with non-engineers. You will work directly with Product Management, Professional Services and, at times, clients.
- You stay current. This field moves quickly. We expect you to keep up without being asked to.

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.

📌 AI Engineer (Mumbai)
🏢 GoldenSource
📍 Mumbai