E
Senior Software Engineer - 3 (Orchestration & Validation)
Exotel Techcom Pvt Ltd
Bengaluru · Karnataka · in
पूर्णकालिक
1h ago
61%
अच्छा
नौकरी विवरण
About Us Exotel is one of Asia’s largest customer communication platforms. We are on a mission to move enterprise customer communication to the cloud. In 2020, we powered over 4 billion calls and connected over 320 million people. We work with some of the most innovative companies such as Ola, Swiggy, Zerodha, Whitehat Jr, Practo, Flipkart, GoJek, etc. We also power customer communication for some of the top banks in the country. Join us on this journey to make a difference to how companies look at customer communication. Read our growth story here. Tech @ Exotel Exotel engineering solves some really cool infrastructure level problems with the goal of ensuring no one misses a call or an SMS. Our focus is on building very fault-tolerant, loosely coupled, scalable and real-time distributed system We are generally agnostic of language, technology or tools. Currently, our tech stack is built on Java, Golang, Node.js, Ruby and PHP. We use Postgres, Citus, Redis, Aerospike, MySQL as data stores, ElasticSearch for search and Kafka, Beanstalkd for queuing. We emphasize a lot on clean abstractions of code, loosely coupled services and good coding practices We are very strong believers in “you built it, you own it!”. And running a distributed system is very different from just building one! We are crazy about high availability. Engineering Team @ Exotel The “Engineering” Team builds our Platform & Enterprise Contact Center solutions enabling customers to cater to their customer support, sales and collection use cases. The team has to cater to diverse requirements from domestic/international customers. Examples include developing voice, and messaging stacks, integrating our contact centre with CRMs such as Salesforce, ServiceNow, Zoho, Freshdesk, etc OR into customers’ own workflows, developing omni channel capabilities for our contact center etc. About the Role With evolving role of a developer as an orchestrator of agents that build software, we are seeking builders who can build and manage agents that validate software built with AI. You will build and train the AI agents that review pull requests and validate every release, read the code yourself to judge what those agents report, and stay the human in the loop who decides whether a change is safe to ship. You will be expected to understand their code and archiecture enough to suggest+validate deep whitebox approaches. You will debate with owners of PRs to convince why some code is either safe or unsafe to ship. What You Will Do Regression Analysis: build and operate agents, understand the surrounding service code, work out the blast radius, and arrive at risks and gaps. Risks identify what could go wrong and gaps identify what was needed vs what was built. Decide whether the AI identified risks and gaps are accurate. Validate static analysis with dynamically built tests agentically as the human in the loop Develop and train the test agents: prompts, skills, tool contracts and the CLI/MCP tools that give agents access to code, GitLab, Jira and test environments. Make agent assertions white-box: decide which DB rows, Redis keys, Kafka events or SIP/RTP traces prove a change works, and teach the agents to check them, not just the UI. Measure the agents: track the valid-bug rate, the false-positive rate and time to signal, and keep a set of past MRs with known outcomes to regression-test the agents whenever a prompt or tool changes. Turn every confirmed finding and every escaped defect into a deterministic regression test in CI. Agents feed the release gate; they do not replace it. Tell environment failures from product failures in ephemeral and pre-prod setups, and fix or escalate at the right layer. Validate AI products, whose outputs vary from run to run, with evaluation datasets and LLM-as-judge checks calibrated against human review. Review the agent and test work of other engineers, and help the team adopt this way of working. What We Look For Must Haves 6–8 years in software engineering with a Bachelor’s/Master’s in Computer Science or equivalent, including substantial time writing production code in a product engineering organisation. Expertise in one of Java / Go / Python, and enough reading fluency in the others to review an PR in them. Working knowledge of Postgres / Citus / MySQL / Redis and Kafka / Beanstalkd — enough to know what state a change should leave behind and how to check it. Experience with REST APIs and microservices in large-scale distributed systems. Contact centre or telephony exposure is a plus: SIP, RTP, WebRTC, TCP/UDP. Hands-on work building or seriously extending an LLM agent workflow (tool use, multi-step plans, retries, evaluation) and shipping it to real use. CI experience: Running in Jenkins / GitLab CI / GitHub Actions. Comfort with Docker and Kubernetes at the level of reading pod logs and debugging a failed environment. Clear judgement on when an agent is wrong, and the habit of proving it with evidence. Good To Have Evaluating LLM outputs: eval datasets, rubrics, LLM-as-judge. Observability tooling (Prometheus, Grafana, ELK, OpenTelemetry). Exposure to AngularJS, React or Node.js front ends. Soft Skills A “DevOps” mindset: you own what you validate, through to production. Excellent written communication — your sign-off and bug reports are read by developers and release managers. Says no to agent verdicts that cannot be backed with evidence. Team player who raises the bar through reviews and mentoring.