Artificial Intelligence Engineer

About the Role:

Hands-on AI Engineer to design, build, and maintain production-grade AI agents that automate workflows across NBFC operations, legal and regulatory research, investment analysis, and corporate functions. You will own the full lifecycle — scoping, architecture, prompt and context design, tool integration, deployment, and iteration. This is a builder role: the output is reliable, production-ready agents, not experiments.

Key Responsibilities:

Design and deploy single-purpose and multi-step AI agents using modern agent frameworks. Architect context windows with deliberate handling of system prompts, retrieval, inter-step context passing, and memory. Author, maintain, and version-control prompts in structured Markdown. Integrate agents with internal systems and external data sources through APIs and MCP. Build evaluation harnesses with test sets, regression checks, and human-in-the-loop review gates. Implement guardrails for hallucination control, primary-source citation, and refusal handling for out-of-scope tasks. Monitor production performance across cost, latency, reliability, and accuracy.

Must-Have:

• Context engineering — Deep understanding of designing context windows for specific use cases; managing long-context degradation; structuring short-term, long-term, and episodic memory; context compression without compromising primary-source fidelity.
• Markdown prompting — Strong structured prompt-writing skills with clear hierarchy, few-shot examples, output-schema enforcement, edge-case handling, and negative examples for failure modes.
• Agent frameworks — Production experience with two or more of LangGraph, CrewAI, AutoGen, Anthropic Agent SDK, or OpenAI Agents API. Strong understanding of tool-calling, MCP, orchestration patterns, and agent workflows.
• Python — Strong proficiency with async workflows, vector databases, document parsing (PDF, DOCX, XLSX, OCR), and production-grade backend development.
• Retrieval systems — Experience with chunking strategies, embeddings, hybrid search, re-ranking pipelines, and citation preservation.
• Model selection — Practical judgement in choosing frontier, mid-tier, and lightweight models based on quality, latency, and cost trade-offs.
• Evaluation discipline — Experience building eval datasets, measuring accuracy and error rates, running regression testing, and conducting A/B evaluations before deployment.

Good-to-Have:

Familiarity with Indian regulatory and case-law data sources such as BSE/NSE, MCA21, SEBI, RBI, IBBI, and Indian legal databases including SCC Online. Exposure to NBFC, securities law, insolvency, or compliance workflows. Awareness of DPDP Act requirements, data residency considerations, and audit-trail expectations for regulated AI systems. Experience with fine-tuning approaches such as LoRA or SFT when prompting reaches limitations. Basic frontend capability for building lightweight internal agent interfaces.

Profile:

3–6 years of software engineering experience, including at least 12–18 months building production-grade LLM agents (not prototypes or demos). Strong engineering background or equivalent practical experience. Portfolio of deployed agents or detailed technical write-ups carries more weight than pedigree. Comfortable operating in fast execution cycles with direct collaboration alongside the Promoter/MD. Strong bias toward shipping, validating outputs, and clearly communicating uncertainty instead of over-claiming.

Compensation:

Market-competitive fixed + performance variable tied to agents shipped and adoption. API budgets, hardware, and tooling provided.