About the role

We help mid-market and mid-tier enterprises move from AI intent to AI in production — with the outcome measured, governed, and owned. We are not a strategy firm and not a development shop; we sit in the gap between them, where most enterprise AI initiatives die.

This is a senior, client-facing role. Your first engagement is with a law firm that wants to understand what AI can realistically do for their practice, which tools are worth adopting, and what needs to be purpose-built. You will be embedded with them — sitting with partners and associates, learning how the work actually gets done, and translating that into a concrete, sequenced plan they can act on.

You will then do the same for clients in banking, insurance, manufacturing and shared-services centres.

This is not an implementation-only role and it is not a research role. You need to be credible in a partner's office and credible at a terminal, in the same week.

What you will actually do

Discovery and opportunity mapping

- Sit inside the client's workflows — matter intake, conflict checks, document review, contract abstraction, due diligence, legal research, drafting, knowledge management, time capture and billing — and understand where the real friction is, as opposed to where people say it is
- Quantify baselines before proposing anything: hours spent, turnaround times, error rates, cost per unit of work. No recommendation goes out without a measured baseline behind it
- Interview across seniority — a senior partner and a second-year associate will describe the same process completely differently, and the gap between those accounts is usually where the opportunity sits

Assessment and prioritisation

- Generate a candidate use-case portfolio and score it on value, feasibility, risk and data readiness
- Kill the majority of candidates and say clearly why. A client who receives twenty "opportunities" has received nothing




- Assess whether the client's data and systems can actually support each candidate — most AI failures are data failures, not model failures

Tool evaluation and build-vs-buy

- Evaluate the legal AI tool landscape on merit rather than marketing — contract review and abstraction platforms, legal research assistants, document intelligence, e-discovery, drafting and knowledge management tools, and general-purpose LLM deployments
- Make a clear build-vs-buy call per use case with the reasoning exposed
- Run structured vendor evaluations: security posture, data residency, model provenance, pricing at scale, exit terms

Solution design

- Write specifications precise enough for an engineering pod to build from — architecture, integration points, evaluation criteria, human-in-the-loop design, failure modes
- Build lightweight prototypes yourself where a working demonstration will settle an argument faster than a document
- Design the evaluation harness and golden dataset up front, not after the build

Governance and risk

- Design for the client's actual constraint set. In a legal practice that means client confidentiality and privilege, conflicts of interest, DPDP obligations, retention and disposal, and the professional consequences of unverified AI output reaching a filing or a client
- Specify audit logging, human review checkpoints, output constraints and escalation paths as part of the solution, never as an afterthought

Adoption

- Design the rollout, run enablement, and stay close enough to see where usage actually lands. A tool nobody opens is not a deployment




- Build the client's internal capability deliberately. Every engagement includes a knowledge transfer plan and a capability assessment at exit — we are explicitly not trying to make ourselves permanent

What we need from you

Essential

- 7+ years total experience, with at least 2 years working substantively with LLMs, RAG systems, agentic workflows or applied ML in a production or near-production context
- Demonstrated consulting or client-facing capability. You have run discovery, presented to senior stakeholders, been challenged, and held your position with evidence. This is weighted as heavily as technical depth
- Genuine hands-on ability. You can build a working prototype, evaluate a model's output rigorously, and read enough code to know when a vendor is overselling. You do not need to be the strongest engineer in the room, but you cannot be dependent on one to have an opinion
- Process analysis instinct. You are drawn to how work actually flows through an organisation, and you notice the gap between the documented process and the real one
- Structured written communication. Our deliverables are documents. You will write assessment reports, solution specifications and board-ready summaries, and they need to be good
- Intellectual honesty. You will regularly need to tell a client that a use case they are excited about will not work, or that they should not do AI in a function yet. Candidates who cannot do this are not usable in this role

Strongly preferred

- Exposure to legal, qualified services, or another document-and-judgement-heavy domain (audit, compliance, insurance underwriting, medical coding)
- Familiarity with India's regulatory environment for AI and data — DPDP, and sectoral expectations in financial services
- Experience evaluating or deploying enterprise SaaS, including security review and vendor negotiation
- Track record of designing evaluation methodology for AI systems

📌 AI Solutions lead (Mumbai)
🏢 63SATS Cybertech
📍 Mumbai