Website:
aivar.tech
Job details:
Platform: AWS + Aivar (cross-platform — Convogent, Velogent, Kubogent)
Level: Senior (5–10 years overall | 2+ years shipping production GenAI/agentic systems)
Location: Bangalore / Coimbatore / Mumbai, with significant time embedded at customer sites (India across multiple Cities)
About The Role
Aivar is an AWS Preferred Partner backed by Bessemer Venture Partners and Sorin Investments. Our customers across fintech, healthcare, and logistics rely on us to take them from AI experimentation to production at scale. The FDE is the role that makes that promise real on the ground.
As a Forward Deployed Engineer (FDE) at Aivar, you are an embedded builder who closes the gap between frontier AI capabilities and production-grade reality inside enterprise customer environments. This is not an advisory role. You operate as a builder-consultant— moving past high-level architecture to code, debug, and jointly ship bespoke agentic solutions directly within the customer's stack, alongside their engineering team.
You unblock production. The integration complexities, data readiness issues, identity and security boundaries, and state-management challenges that keep AI stuck at "interesting demo" — those are your problem to solve. You embed with strategic customers and serve a dual purpose: providing **white-glove deployment** of Aivar's accelerator platforms (Convogent for voice AI, Velogent for governed agentic automation, Kubogent for Kubernetes-native AIOps), and acting as a **critical feedback loop** — translating real-world field insights into Aivar's product roadmap.
Responsibilities
- Serve as the Senior developer for complex AI applications inside strategic customer accounts — taking projects from rapid prototype to production-grade agentic workflows (multi-agent systems, MCP servers, RAG pipelines, governed automation) that deliver measurable return on investment
- Architect and code the connective tissue between Aivar's AI accelerators, AWS AI services, and the customer's live infrastructure — APIs, legacy data silos, identity systems, security perimeters, and existing enterprise applications
- Build high-performance evaluation (Eval) pipelines and observability frameworks to ensure agentic systems meet requirements for accuracy, safety, latency, and cost
- Identify repeatable field patterns and technical friction points in Aivar's AI stack — convert them into reusable modules, internal libraries, or formal product feature requests for the Engineering teams
- Co-build with customer engineering teams — instill Aivar-grade development practices, ensuring long-term project success and high end-user adoption after Aivar's direct involvement winds down
- Own the technical relationship with customer engineering leadership — translate executive intent into shipped systems and ship the engineering credibility that earns expansion deals
- Anchor production cutover and post-go-live stabilization for the systems you build
Must-Have Requirements
Technical Skills
- 3+ years hands-on Python plus relevant ML packages (Hugging Face Transformers, Keras / PyTorch, NumPy / Pandas) — production-grade engineering, not notebook-only
- Applied AI experience building systems around pretrained models— prompt engineering, fine-tuning, Retrieval-Augmented Generation (RAG), and orchestrating model interactions with external tools to deliver real solutions
- Multi-agent systems** experience — using frameworks like LangGraph, CrewAI, Strands Agents, AutoGen, or Bedrock Agents — and the underlying patterns: ReAct, self-reflection, hierarchical delegation, tool use, structured output
- AWS-native AI delivery — Bedrock (Anthropic, Llama, Titan, Mistral families), SageMaker** for training/hosting, Lambda, Step Functions, OpenSearch / pgvector for retrieval, S3, IAM
- MCP (Model Context Protocol) — understanding of MCP server patterns and tool-server design
- Evaluation engineering — building eval datasets, judge-model patterns, regression gates, drift monitoring; familiarity with frameworks like Ragas, DeepEval, Promptfoo, or custom harnesses
- Systems design— ability to architect and explain data pipelines, ML pipelines, and ML training and serving approaches end-to-end
- Bachelor's degree in Science, Technology, Engineering, Mathematics, or equivalent practical experience
Domain / Business Experience
- Direct experience **working with enterprise customers in a technical capacity** — translating ambiguous business problems into concrete AI systems
- Track record of shipping AI features into **production**, not just POCs or demos
- Comfort operating inside customer security and compliance constraints (IAM, data residency, audit logging, change control)
Mindset & Culture
- Action-oriented— relentless focus on solving the customer's problem and getting code into production
- Bias to ship — operates in weeks-not-months cycles; iterates rather than perfecting
- Daily user of AI coding tools (Claude, Copilot, Cursor, or equivalent) — treats AI fluency as a multiplier on personal output
- High customer-facing maturity — can sit in a room with customer CTO/VP Engineering and ship code with their team in the same week
- Founder-mentality — owns the outcome, not just the task; comfortable with ambiguity
Nice-to-Have
- Master's degree / Phd in Computer Science, Engineering, or a related technical field
- Experience training and fine-tuning models in large-scale environments (image, language, recommendation) with GPU/TPU accelerators
- Knowledge of LLM-native metrics — tokens/sec, cost-per-request, time-to-first-token, p95 latency — and techniques for optimizing them
- Hands-on experience with state management in long-running agentic workflows and granular tracing (Langfuse, LangSmith, OpenTelemetry, AWS X-Ray)
- Experience in regulated verticals — Fintech, healthcare, logistics— and the compliance overlays they bring (RBI, HIPAA, SOC 2, DPDP Act)
- Open-source contributions, technical writing, or speaking on agentic AI / LLM systems
- Voice AI experience (Twilio, telephony) — relevant for Convogent customers
- Kubernetes / EKS / MLOps depth — relevant for Kubogent customers
- Prior FDE / Solutions Architect / Customer Engineering experience at a hyperscaler or AI platform company
## Key Technologies
Python, AWS Bedrock, SageMaker, Lambda, Step Functions, OpenSearch / pgvector / Pinecone, LangChain / LangGraph / CrewAI / Strands Agents / Bedrock Agents, MCP, Hugging Face Transformers, FastAPI, Pytest, Ragas / DeepEval / Promptfoo, Langfuse / LangSmith, OpenTelemetry, CloudWatch, X-Ray, Docker, Terraform / CDK basics, GitHub Actions / CodePipeline
What Success Looks Like (First 90 Days)
- Embedded with at least one strategic customer account and operating as a trusted member of their engineering team
- First agentic workflow shipped to production inside that customer's environment — with eval gates, observability, and a clean handover plan
- Connective-tissue components built — integration into the customer's auth, data, and security boundary
- First field-pattern insight captured and pushed back to Aivar Engineering as a reusable module or product feature request
- Documentation pack complete — LLDs, runbooks, eval baselines — sufficient for the customer team to operate the system
Why This Role Is Different
Most "AI engineer" roles either sit in an internal product org and never see a customer, or sit in a consulting role and never get to ship production code. The FDE role is the rare combination: you are an **engineer first**, but you build inside the messiest, highest-stakes context that exists — a real enterprise customer trying to put AI into production. You'll see what actually breaks in the field, fix it in code, and that fix will often become the next thing the product team ships to everyone else.
If you've ever felt that AI products are too far from where the real problems live, this role exists to close that gap.
Click on Apply to know more.