ABOUT THE ROLE
The Data + Automation team at nCircle Tech builds the data and automation infrastructure behind decision
making across the business. Our platform runs on Databricks and AWS. We're hiring a solid software
engineer to design, build, and ship production services — including LLM-powered agents and the tooling
around them. This is a hands-on, code-first role.

WHAT YOU'LL WORK ON
● Designing, building, and productionizing LLM agents: tool-use loops, retrieval, orchestration,
evaluation, and guardrails
● Writing the production-grade services and APIs around those agents in Python and Node/TypeScript
● Deploying and operating on AWS — provisioning with IaC, wiring CI/CD, owning your code from
commit to production
● Contributing to a shared codebase with feature branches, code review, and Conventional Commits

REQUIRED
Software Engineering (the core)
● Strong Python and Node/TypeScript — writes clean, tested, production-grade code, not scripts
● Solid fundamentals: data structures, API design, error handling, testing, debugging real systems
● Feature branches and PRs; comfortable both giving and receiving code review; meaningful commit
messages
● Follows engineering standards (12-factor, testing, documentation) without hand-holding

AWS + IaC
● Real AWS experience — the environment our services run on
● Deploys with Infrastructure as Code (Terraform, CDK, or equivalent) rather than clicking through
consoles
● Comfortable with CI/CD, environment promotion, and AWS basics (S3, IAM, networking)

LLM Agents (non-negotiable)
This is a hard requirement, not a nice-to-have. We are a small team and cannot train this skill on the job
— you must already know how to build agents and be able to contribute to real work quickly.




● Has built and shipped LLM agents to production — not demos or prototypes; systems that ran and
were relied on
● Can explain the agent loop from experience: how the model decides to call a tool, observes the result,
and re-decides across multiple steps

Senior Software Engineer, AI Agents
● Has designed retrieval (RAG) end-to-end — chunking, indexing, hybrid or re-ranked retrieval — and
can justify each choice
● Has a real evaluation story: how they measured whether the agent was correct, and the worst failure
mode they hit in production and fixed
● Knows when NOT to use an agent — where a plain LLM call or deterministic code is the better answer
● Framework-agnostic — LangChain, CrewAI, Databricks Agent Framework, or raw tool-use loops all
fine; what matters is that you've shipped real agents

Databricks (working proficiency)
Our platform runs on Databricks, so this is required — but working proficiency, not mastery. You should
be able to build and ship in Databricks unaided; you don't need to be a Spark-tuning specialist.
● Comfortable building and running Databricks notebooks and Jobs (task chains, parameters,
schedules) end-to-end
● Understands the Unity Catalog three-part namespace (catalog.schema.table) and works within it
● Has written Delta Lake operations beyond basic reads — e.g. MERGE, overwrite,



schema evolution
● Can navigate and debug an existing pipeline, not only write greenfield code

GOOD TO HAVE
● Deeper Databricks / Spark depth — performance tuning, advanced Unity Catalog governance
● MCP (Model Context Protocol) exposure
● Agent observability and tracing tools in production
● Construction, AEC, or project-based industry experience
● Jira

GENERAL BAR
● 5+ years writing code, not configuring tools
● Ships independently: scopes work, executes, communicates progress
● Can ramp on an unfamiliar system by reading docs and asking targeted questions

BenefitsSkills / responsibilities the panel was screening for

- Databricks platform depth: Unity Catalog (governance, lineage, three-level namespace, external/managed tables, row/column-level masking), Delta Lake, cluster management (all-purpose vs. job vs. serverless compute), Databricks Workflows

- Data modeling: medallion architecture (bronze/silver/gold), SCD Type 1/2, surrogate key generation, star vs. snowflake schema design

- Data quality & governance: schema evolution/enforcement, null handling, data contracts, naming standards/glossaries, tools like DQX, Great Expectations, Atlan

- Pipeline engineering: idempotency, retry/failure handling, deletion propagation (soft/hard deletes)

- Emerging tool fluency: interest in AI-assisted data engineering (Claude skills/plugins for profiling reports, dynamic dashboards) — they're actively moving in this direction

- Business/domain awareness: understanding how a construction company (multi-domain: HR, risk, project delivery, competitive intelligence) uses data for things like cost prediction and bid-win-probability modeling

📌 Data Engineer (Pune)
🏢 nCircle tech
📍 Pune