Senior Consultant | Backend Developer | Mumbai and Gurugram

Deloitte Shared Services India; Gurugram District, Mumbai; Presencial; INR 2.000.000 - 3.000.000; Jornada completa

Descripción de la vacante

Deloitte Shared Services India is seeking an experienced Agentic App Engineer to lead the design and development of agentic and intelligent systems with strong ownership of core algorithms, model behavior, and decision logic. You will minimize reliance on managed AI services and abstraction-heavy frameworks while driving architectural discipline.

The role requires architecting end-to-end agentic systems with control over model internals, reasoning flows, and execution pipelines.

Formación

• 4 to 7 years of strong experience building AI-driven systems with ownership of model behaviour and algorithms.
• Deep understanding of agentic architectures, multi-agent coordination, autonomous decision-making systems, and custom orchestration approaches.
• Ability to design agent frameworks from scratch or significantly extend existing ones rather than relying on out-of-the-box abstractions.
• Strong understanding of LLM internals, transformer architectures, inference behavior, tokenization, attention, trade-offs.
• Experience with open-source LLMs (e.g., LLaMA-family, Mistral, Mixtral, Falcon, Qwen) and self-hosted deployments.
• Advanced reasoning strategies: chain-of-thought alternatives, tree/graph-based reasoning, planners, criticagent loops, tool reasoning.
• Prompting as a controlled technique, not the primary abstraction.
• Experience with LangChain, LangGraph, AutoGen, CrewAI as reference tools, with ability to bypass or replace them with custom implementations.
• Preference for code-first, algorithm-driven approaches over declarative or black-box pipelines.
• Deep experience in RAG pipelines, embeddings, vector similarity search, hybrid retrieval, reranking models, and knowledge grounding.
• Strong grasp of information retrieval, indexing, and retrieval algorithms.
• Strong systems engineering skills with Python and/or Java/Go for AI backends.
• Experience deploying AI workloads on GCP, including self-managed inference, GPUs, containers, and scaling strategies.
• Familiarity with MLOps practices: model versioning, evaluation pipelines, controlled rollouts, experimentation.
• Deep experience in AI observability: tracing, latency profiling, quality evaluation, cost monitoring.
• Strong understanding of AI security, data privacy, access control, and compliance; preference for in-house controls over managed tooling.
• General Insurance domain knowledge is a strong advantage.

Responsabilidades

• Lead design and development of agentic and intelligent systems with ownership of core algorithms, model behavior, and decision logic.
• Architect end-to-end agentic systems with deep control over model internals, reasoning flows, tool-use logic, and execution pipelines rather than wrapper-based orchestration.
• Own algorithmic and model-level aspects of agent behavior, including reasoning strategies, planning, fallback logic, and failure handling.
• Design and implement custom agent orchestration, memory systems, context management, and stateful workflows without over-dependence on abstractions.
• Build and fine-tune open-source and self-hosted LLMs or hybrid approaches based on performance, cost, and security trade-offs.
• Develop custom RAG pipelines, embeddings, ranking, re-ranking, and grounding mechanisms tailored to enterprise knowledge systems.
• Integrate agentic solutions with enterprise applications, APIs, data platforms, and internal ML systems.
• Establish low-level design patterns, reusable libraries, and internal frameworks for building agent-based systems from first principles.
• Optimize model inference, reasoning efficiency, latency, cost, and response quality, including prompt-free and programmatic reasoning approaches.
• Drive Responsible AI practices with guardrails, explainability, auditability, and governance controls.
• Collaborate with business stakeholders to translate complex requirements into deterministic, reliable agent-driven solutions.
• Lead production support, deep debugging, RCA, and performance tuning at both system and model levels.
• Mentor junior consultants on core AI concepts, algorithms, and system design.

Conocimientos

AI system design

Python/Java/Go

LLM internals

LangChain/LangGraph usage

RAG pipelines

Self-hosted LLMs

Educación

BE / B.Tech in Computer Science or related

Herramientas

LangChain

LangGraph

AutoGen

CrewAI

Descripción del empleo

Agentic App Engineer

Roles & Responsibilities :

• Lead the design and development of agentic and intelligent systems with strong ownership of core algorithms, model behavior, and decision logic, minimizing reliance on managed AI services and abstraction-heavy frameworks.
• Architect end-to-end agentic systems with deep control over model internals, reasoning flows, tool-use logic, and execution pipelines rather than wrapper-based orchestration.
• Take algorithmic and model-level ownership of agent behavior, including reasoning strategies, planning algorithms, fallback logic, and failure handling.
• Design and implement custom agent orchestration, memory systems, context management, and stateful workflows without over-dependence on LangChain/LangGraph-style abstractions.
• Build and fine-tune open-source and self-hosted LLMs or hybrid approaches (open-weight + proprietary) based on performance, cost, and security trade-offs.
• Develop custom RAG pipelines, embedding strategies, ranking, re-ranking, and grounding mechanisms tailored to enterprise knowledge systems.
• Integrate agentic solutions with enterprise applications, APIs, data platforms, and internal ML systems.
• Establish low-level design patterns, reusable libraries, and internal frameworks for building agent-based systems from first principles.
• Optimize model inference, reasoning efficiency, latency, cost, and response quality, including prompt-free and programmatic reasoning approaches where applicable.
• Drive Responsible AI practices with in-house guardrails, explainability techniques, auditability, and governance controls.
• Collaborate with business stakeholders to translate complex requirements into deterministic, reliable agent-driven solutions, avoiding black-box dependencies.
• Lead production support, deep debugging, RCA, and performance tuning at both system and model levels.
• Mentor junior consultants on core AI concepts, algorithms, and system design, not just framework usage.

Education

• BE / B.Tech in Computer Science, Engineering, or a related discipline.

Required Experience & Skills.

• 4 to 7 years of strong experience building AI-driven systems, with demonstrated ownership of model behaviour and algorithms, not just API integration.
• Deep understanding of agentic architectures, multi-agent coordination, autonomous decision-making systems, and custom orchestration approaches.
• Ability to design agent frameworks from scratch or significantly extend existing ones rather than relying out-of-the-box abstractions.
• Strong understanding of LLM internals, transformer architectures, inference behavior, tokenization, attention, trade-offs.
• Experience with open-source LLMs (e.g., LLaMA-family, Mistral, Mixtral, Falcon, Qwen) and self-hosted deployments.
• Advanced reasoning strategies: chain-of-thought alternatives, tree/graph-based reasoning, planners, criticagent loops, tool reasoning.
• Prompting as a controlled technique, not the primary abstraction.
• Experience with LangChain, LangGraph, AutoGen, CrewAI as reference tools, with proven ability to bypass, customize, or replace them with custom implementations.
• Preference for code-first, algorithm-driven approaches over declarative or black-box pipelines.
• Deep experience in RAG pipelines, embeddings, vector similarity search, hybrid retrieval, reranking models, and knowledge grounding.
• Strong grasp of information retrieval, indexing, and retrieval algorithms.
• Strong systems engineering skills with Python and/or Java/Go for AI backends.
• Experience deploying AI workloads on GCP, including self-managed inference, GPUs, containers, and scaling strategies.
• Familiarity with MLOps practices: model versioning, evaluation pipelines, controlled rollouts, and experimentation.
• Deep experience in AI observability: tracing reasoning paths, latency profiling, quality evaluation, and cost monitoring.
• Strong understanding of AI security, data privacy, access control, and compliance, with preference for in-house controls over managed tooling.
• General Insurance (Health, Motor, Travel) domain knowledge is a strong advantage.

Soft Skills

• Demonstrates strong ownership mindset, curiosity at the algorithmic level, and comfort working with ambiguity and complex systems.
• Collaborates effectively across engineering, data science, and business teams, clearly explaining technical trade-offs and design decisions.
• Exhibits professional integrity, reliability, and clear, respectful communication with stakeholders.