Open to opportunities

AkashNikam

Building production agentic RAG systems and MCP-integrated developer tools at John Deere - from a plant-floor maintenance copilot to a multi-agent platform on AWS Bedrock.

View my work ↓GitHub ↗

Years experience

Production AI systems shipped

268ms

P95 RAG retrieval latency

40+

MCP tools shipped

agent.py

🤖

3 AI Systems

shipped at Deere

⚡

268ms P95

RAG latency

🔗

40+ Tools

via MCP

🌐

EKS

production deploy

About me

I'm an AI engineer who builds agentic systems for a living. Currently a Senior Software Engineer at John Deere, where I design and ship multi-agent platforms that change how engineering teams actually work - the intersection of LLMs, backend systems, and shipping real things people use.

The thing that drives me: I learn by building. When I want to understand something - OAuth and webhooks, RAG retrieval, bounded AI pipelines - I don't just read about it, I ship a working version. That's where ZapBridge, ARKEN, and PlantIQ came from: each one started as a question I couldn't answer from a blog post.

I'm allergic to hand-waving. I'd rather say "I haven't built that yet" and go build it than fake my way through - and when I build, I care about the why behind every decision: timing-safe validation, idempotency, graceful degradation, not just code that happens to run.

Before Deere I spent 2+ years at Mavenir building Go microservices for telecom-grade analytics on Kubernetes. Outside of code: dissecting new architectures and occasionally writing on Medium about things I figured out the hard way.

Pune, IndiaOpen to remoteAI EngineerAgentic Systems

Skills & Stack

What I work with

Technologies I reach for when shipping real systems, not just listed but actually used.

AI & Agents

Agentic AIAgentic RAGMulti-Agent SystemsRAGMCPLLM-as-Judge EvaluationLangGraphLangChainAWS BedrockHybrid Search (BM25 + RRF)Vector Search (Qdrant / OpenSearch)Context & Prompt EngineeringAI Agent Governance

Backend

GoPython / FastAPIgRPCProtocol BuffersREST APIsMicroservices

Frontend

TypeScriptReact + ViteJavaScriptWebRTCSSE / Streaming

Databases

PostgresMongoDBRedisDynamoDBCouchDBSinglestore

DevOps & Infra

DockerKubernetesNginxCI/CDGitHub Actions

Experience

Where I've worked

5+ years building AI agent systems and the production backends behind them.

Full profile ↗

John Deere

Senior Software Engineer - AI Agents & Backend

Oct 2023 - Present · Pune, Hybrid

EmbAI-ASCENT

Built EmbAI-ASCENT, a governed VS Code extension for the engineering SDLC: agent-pack distribution via a custom CLI (Agent Package Manager), 3 MCP servers (Azure DevOps, Confluence, GitHub Enterprise), and 4 custom LM tools wired into Copilot Chat.
Designed a 13-phase context-engineering webview with keyword-scoring context assembly and workspaceState-backed session continuity, so engineers can resume multi-day planning threads without losing context.

PlantIQ (Plant Brain)

Built PlantIQ (Plant Brain), an access-controlled RAG copilot for plant-floor maintenance technicians: hybrid BM25 + dense (bge-small-en-v1.5) retrieval fused via Reciprocal Rank Fusion over Docling-chunked SOPs, serving at P95 268ms.
Hardened PlantIQ with deterministic fail-closed guardrails (confidence gating, 15% grounding threshold), JWT-derived plant_id/org_id access scoping, and a CI-gated evaluation harness (dual LLM-as-Judge, Claude Opus + GPT-4o, 0.20 divergence flag).

AI Developer Assistant Platform

Migrated a single-PDF chatbot into the AI Developer Assistant Platform: a ReAct-based agentic RAG system on AWS Bedrock across 7 Lambda-backed knowledge bases (Titan V2 embeddings, OpenSearch Serverless).
Built a FalkorDB code knowledge graph over 9 repositories (tree-sitter, 25 languages) and a multi-agent orchestration layer (Supervisor/Researcher/Coder/Writer) with a sequential RFI pipeline.
Shipped an MCP server exposing 40+ tools / 140+ skills across Slack, CLI, web, and n8n, deployed on EKS with GitOps-driven CI/CD.

Mavenir

Member of Technical Staff I (R&D)

Mar 2022 - Oct 2023 · 1 yr 8 mos · Remote

Built a production Go service for a high-frequency analytics notification system - polls CouchDB every 100ms, matches epoch-time triggers, dispatches notifications, and auto-expires entries after 30 minutes. Used goroutines and channels for concurrent, non-blocking delivery at scale.
Used SingleStore for distributed data storage and document management across Go microservices in production - hands-on with its query model, distributed data patterns, and performance characteristics in a real-time service architecture. Also worked with CouchDB for document management.
Designed and implemented versioned OpenAPI specifications in Go for a microservice platform, making APIs fully configurable for future version changes and significantly reducing migration effort.
Contributed to NWDAF application development on a microservice architecture in Go - service-to-service APIs, distributed state management, and production deployment across Kubernetes-based infrastructure.
Worked with Go, Docker, and Kubernetes throughout R&D - containerized services, K8s deployment patterns, and service-mesh communication in a production microservice environment.
Collaborated cross-functionally with R&D teams to design, build, and ship Go services handling concurrent workloads, real-time data pipelines, and distributed notification systems at telecom scale.

Mavenir

Graduate Engineer

Mar 2021 - Feb 2022 · 1 yr · Remote

Worked hands-on with Go, Docker, and Kubernetes in a production microservice architecture - containerized service deployment, K8s service-mesh patterns, and inter-service communication across distributed components.
Wrote unit tests across 2-3 microservices, improving coverage and reliability of core services - disciplined about testing and regression prevention from day one.
Resolved bugs on the Prediction Application by triaging issues, identifying root causes, and shipping fixes under production constraints - building strong debugging instincts across distributed Go services.
Used CouchDB for document management alongside Go microservices in a distributed data architecture.