Innovapptives Connected Worker Platform is expanding its AI capability from foundational features into a broad portfolio of product-facing AI agents purpose-built for industrial field operations. These agents span maintenance planning, work order automation, safety compliance, operator rounds, and knowledge assistance, all grounded in customer-specific asset data and SOPs.

This role leads the AI Engineering team responsible for designing, building, and operating that agent portfolio in production.

You own the full lifecycle : from architecture and prompt engineering through evaluation, deployment, and reliability. You work closely with Product, Platform, and customer-facing teams to translate industrial use cases into AI capabilities that enterprise customers trust.

What You Own :

- AI Engineering team : 6-8 engineers across agent development, LLM infrastructure, and model evaluation.
- End-to-end agent lifecycle : requirements through architecture, build, evaluation, deployment, and production monitoring.
- RAG and knowledge infrastructure : document ingestion pipelines, chunking strategies, embedding, vector search, and knowledge graph grounding.
- LLM governance : model selection, prompt versioning, bias testing, audit logs, and human-in-the-loop controls. All inference within Innovapptives AWS VPC no data to external LLM endpoints.
- Agent quality : evaluation frameworks, accuracy benchmarks, hallucination monitoring, and output labelling pipelines.
- Sprint delivery and production reliability. Weekly quality scorecard.
- Hiring, performance management, and coaching. Build the team to full operating capacity.

You Must Have :

- 7+ years in software engineering with 3+ years managing teams delivering AI/ML or LLM-powered products in enterprise production.
- Hands-on experience with LLM orchestration frameworks (LangGraph, LangChain, or equivalent) and multi-step agentic workflows.
- Strong grasp of RAG architecture : document pipelines, chunking, embedding, vector databases, re-ranking,



and similarity thresholds.
- Experience with managed inference infrastructure : AWS Bedrock, SageMaker, or equivalent.
- Track record shipping AI product features on schedule in a SaaS context not just prototypes or internal tools.
- Familiarity with AI observability : prompt tracing, hallucination detection, and output evaluation (Langfuse, Ragas, or equivalent).
- Data-driven : model evaluation scores, accuracy/recall metrics, agent success rates, and DORA metrics for the team.
- Robust engineering standards : prompt discipline, eval-driven development, responsible AI controls, and production-grade reliability.

Nice to Have :

- Knowledge graph architectures (AWS Neptune, Neo4j) for grounding agent outputs in structured asset data.
- Industrial domain knowledge : EAM, ERP integrations (SAP, Maximo), maintenance workflows, or field operations.
- Multi-agent orchestration patterns : tool calling, agent-to-agent delegation, and human-in-the-loop checkpoints.
- Vision models or multimodal AI : image-based defect detection, document OCR, or form digitisation.
- MLOps and LLMOps : model versioning, A/B evaluation, and continuous prompt optimisation pipelines.
- Cloud cost optimisation for LLM workloads : token budgets, model tiering, and caching strategies.
- MongoDB and change stream-based event architectures.

You Will Be Measured On :

- Agent production reliability ?99% uptime. Zero silent failures in retrieval or inference pipelines.
- On-time feature delivery ?90%. Regression rate <10%.
- Agent accuracy : defined evaluation benchmarks met for each agent before GA release.
- Customer adoption :



agents in active use by ?3 enterprise customers within first two quarters.
- LLM governance coverage : every production AI feature has audit logs, human-in-the-loop controls, and bias test results on file.
- Team build-out : AI Engineering operating at full 68 HC with clear ownership and on-call rotations within 90 days.

Tech Stack & Tools :

AI / ML :

- AWS Bedrock, SageMaker, LiteLLM, LangGraph, Milvus (vector DB), AWS Neptune (knowledge graph), Langfuse

Backend :

- Node.js / TypeScript, Python, MongoDB

Infrastructure :

- AWS, Docker, GitLab CI/CD

Observability :

- Langfuse, Sentry, CloudWatch

Tools :

- GitLab, Jira, SonarQube

Compensation & Growth :

- Reports to VP PE&A.; Path to Sr. EM or platform leadership as integration becomes a core horizontal capability.

What We Offer :

- Competitive compensation and equity tied to measurable impact on AI accuracy and performance.
- A platform to shape the semantic intelligence layer of a category-defining industrial SaaS company.
- Access to cutting-edge AI, data, and observability toolchains for continuous learning and innovation.

Innovapptive does not accept and will not review unsolicited resumes from search firms. Innovapptive is an equal opportunity employer and is committed to a diverse and inclusive workplace. Qualified applicants will receive consideration for employment without regard to race, color, religion or creed, alienage or citizenship status, political affiliation, marital or partnership status, age, national origin, ancestry, physical or mental disability, medical condition, veteran status, gender, gender identity, pregnancy, childbirth (or related medical conditions), sex, sexual orientation, sexual and other reproductive health decisions, genetic disorder, genetic predisposition, carrier status, military status, familial status, or domestic violence victim status and any other basis protected under federal, state, or local laws.

📌 Innovapptive - Engineering Manager - AI (India)
🏢 Innovapptive
📍 India