New Delhi, India
Senior Software Engineer (SDE-3) · Softify Business Cloud
Senior Software Engineer with 6+ years building high-performance distributed systems and scalable full-stack applications.
AI work
RAG
Production LangChain RAG pipelines and LLM microservices on Kubernetes at Softify — querying business data and summarizing operational logs for on-call triage. LangGraph-based agent workflows, LangSmith tracing and evaluation, vector search, structured outputs, OpenAI/Claude APIs, and LLM observability on multi-tenant containerized infra with real-time MQ streaming.
About
Senior Software Engineer with 6+ years of experience in end-to-end design, architecture, and implementation of high-performance distributed systems and scalable full-stack applications.
I own delivery end-to-end across Go/Node.js backends, React/TypeScript frontends, and Kubernetes-backed microservices — including a low-latency trading platform (10K TPS, P99 < 25ms), a six-service trucking ERP, and production LangChain RAG / LLM services. I already operate at senior IC scope today — sprint planning, code reviews, technical mentorship — and spent four years fully remote at Cheers Interactive and NeuralSift before my current SDE-3 role at Softify.
Core strengths:
I'm deliberately pushing further into AI — I've shipped RAG and LLM microservices in production, and I'm building agentic workflows (tool calling, orchestration, structured outputs) embedded in how software gets built and operated.
Open to senior full-stack roles — remote or hybrid in Gurugram (Delhi NCR).
Work
Education
Skills
My Projects
From low-latency trading backends and ERP microservices to production AI agents — here's a selection of the systems I've built.
Ingests noisy service logs and returns a structured incident brief — severity, affected services, likely root cause, recommended actions, customer impact — using the OpenAI SDK with a Zod schema enforced so downstream alerting and dashboards always get predictable JSON. Same pattern as the production LLM ops-summarization service.
Anthropic Claude SDK agent that reads an inbound support ticket and autonomously looks up the customer SLA/plan, searches known issues, creates an escalation for the right team, and drafts a customer-facing reply. The model chooses which typed tools to call based on ticket context — agentic routing, not a fixed workflow.
Anthropic Claude SDK agent that takes a git diff and returns a one-line summary, risk level, findings with severity + fix suggestions, and an approve / request-changes decision — validated with Zod so CI quality gates never break on malformed JSON.
Multi-agent orchestration with the OpenAI SDK: a planner breaks the question into research steps, a researcher gathers sources, and a synthesizer returns one actionable recommendation — each sub-agent with a narrow job and a typed input/output contract.
Supervisor + subagents built on the Vercel AI SDK: a ToolLoopAgent commander fans out to specialist agents (log-analyst, deploy-auditor, comms-writer) exposed as tools — the model decides who to call and how many times. Parallel specialist execution, schema-validated final report, abort-aware cancellation.
In-portfolio runner for all five agents: streaming traces, loop iterations, and mock or live mode. Paste a DeepSeek, OpenAI, or Anthropic key in the tab — it is never stored on the server.
Go/gRPC trading execution platform at Softify: order ingress, multi-connector fan-out/fan-in, reconciliation, and audit trail. Goroutine pools for parallel connector calls, Redis for hot order state and idempotent dedup, MongoDB trade/audit log — ~10,000 TPS with P99 under 25ms.
Greenfield decomposition of logistics software into six Go microservices (trip orchestration, route planning, billing, delivery events, metrics, notifications) over gRPC on Kubernetes — MongoDB system of record, Redis cache/pub-sub, React ops dashboard. Append-only trip event log + state machine for billing replay, optimistic locking, and POD attestation before invoice finalization.
Production LangChain RAG pipelines and LLM microservices on Kubernetes at Softify — querying business data and summarizing operational logs for on-call triage. LangGraph-based agent workflows, LangSmith tracing and evaluation, vector search, structured outputs, OpenAI/Claude APIs, and LLM observability on multi-tenant containerized infra with real-time MQ streaming.
Combines ANPR with React in a containerized, offline-first environment — a one-stop solution for site and camera settings, traffic violations, detection results, and aggregation. Built with React service workers and Electron.js for desktop support. Uses WebSockets (Pusher client) for real-time alerts and REST APIs for event collection.
Enterprise market-research software at Cheers Interactive: distributed processing for thousands of concurrent survey sessions and long-running batch jobs. Redis distributed locks with unique process tokens, idempotent MQ workers, WebSocket live-session rehydration, and React/TypeScript dashboards tuned for heavy data.
Want to chat? Reach out on LinkedIn or email me at aditya190798@gmail.com.