AI engineer (MCP integration)
Role Summary
We are looking for an MCP Integration Specialist to own how our AI systems connect to third-party services — issue trackers, repositories, CI/CD, test tooling, observability, and ticketing — through the Model Context Protocol (MCP). This is a hands-on integration engineering role focused on safe, governed, production-grade connectivity to external tools.
You will build and operate the integration layer end to end: an MCP gateway that terminates authentication and enforces policy on every tool call, vendor and custom MCP servers, and the credential and trust controls that keep external connections secure. You treat every external service as untrusted by default and design for least privilege. We are looking for a candidate with a strong academic record and solid computer-science fundamentals (data structures, algorithms, and problem-solving) alongside hands-on integration experience.
Key Responsibilities
MCP Gateway & Secure Egress
- Build and operate an MCP gateway as the single egress for all agent tool calls: authentication termination, policy evaluation per call, signature pinning, per-tenant credential injection, and tracing + immutable audit logging.
- Implement OAuth 2.1 flows (PKCE, RFC 9728 / RFC 8414 discovery, Dynamic Client Registration) and move credentials into a secrets manager / vault.
- Standardise on Streamable HTTP transport; reject insecure or deprecated transports; keep audit entries atomic with each outbound call.
Third-Party Service Integration
- Wire vendor-official MCP servers to the right consumers — e.g. Atlassian (Jira/Confluence), Azure DevOps, GitHub/GitLab, Linear, Notion for project & code; Playwright, Postman, SonarQube, Snyk, Semgrep for test & quality; Kubernetes, Terraform, ArgoCD, PagerDuty, Slack for deploy & alerting.
- Map each integration to the data it reads and writes, and enforce allow-listed tool surfaces and human-in-the-loop approval on write operations.
- Run connection tests and validate reads/writes end to end against different tool stacks.
Building MCP Servers
- Author custom MCP servers to fill ecosystem gaps using a shared server SDK/scaffold, with clean typed tool schemas and safe write semantics.
- Fork and harden community servers where vendor-official options do not exist.
Integration Governance & Security
- Maintain a curated registry of integrations with trust tiers (vendor-official / community / custom-built) and a per-tenant connection store (allow-list, approval flags, redacted auth summary, last-test result).
- Hold a hard security posture: treat all servers as untrusted, vet against canonical registries, run vulnerability/signature scanning, and author policy-as-code (OPA / Rego) for tool-call authorisation.
- Stay current on integration and supply-chain risks (tool poisoning, prompt injection, command injection, dependency compromise); maintain integration docs and runbooks.
Collaboration
- Partner with AI/application engineers (the consumers of these integrations) and platform engineers (secrets, networking, gateway deployment).
- Engage external vendor teams on support and roadmap where needed.
- Contribute to engineering standards and delivery cadence.
Required Technical Skills
Domain
Skills & Technologies
Must / Preferred
CS Fundamentals & DSA
Data structures, algorithms, complexity analysis, strong problem-solving
Must
MCP
Model Context Protocol — clients & servers, Streamable HTTP, resources/tools
Must
API Integration
Designing & consuming REST/JSON APIs; webhooks; schema/contract validation; rate limits & retries
Must
Auth & Identity
OAuth 2.1 (PKCE, DCR), RFC 9728 / RFC 8414 discovery, token lifecycle, API-key hardening
Must
Programming
Python 3.10+; Go for gateway / edge services
Must
Policy & Secrets
OPA / Rego policy-as-code; HashiCorp Vault (or equivalent) for credential injection
Must
Integration Security
Signature/vulnerability scanning, untrusted-server posture, prompt-injection & tool-poisoning awareness
Must
Vendor Tooling
GitHub, Atlassian (Jira/Confluence), Azure DevOps, Linear, Slack, PagerDuty, SonarQube, Snyk, Postman, etc.
Must
Observability
OpenTelemetry spans + audit logging for every tool call
Preferred
Platform
Kubernetes deployment of gateway & servers, containerisation
Preferred
Server Authoring
Building MCP servers; typed tool schemas, safe write semantics
Preferred
Agentic Context
Familiarity with LLM agents as MCP clients; tool-calling patterns
Preferred
Qualifications & Certifications
- Strong academic record — B.Tech / B.E. / M.Tech / MCA in Computer Science or a related field from a reputable institution (or equivalent).
- 2–3 years in software / integration engineering, with hands-on third-party API and systems-integration work.
- Strong data structures, algorithms, and problem-solving skills.
- Hands-on MCP experience (building or integrating MCP servers/clients) is a strong plus; otherwise solid API-integration depth with eagerness to ramp on MCP.
Preferred Certifications
- OAuth 2.0 / OpenID Connect or relevant identity/security certification
- Any recognised cloud certification (Azure / AWS / GCP) is a plus
Soft Skills & Cultural Fit
- Security-first instinct — assumes external services are hostile until proven otherwise.
- Strong analytical and debugging skills across system boundaries (auth, networking, schema mismatches).
- Clear written communication — produces precise integration docs and runbooks.
- Collaborative; comfortable engaging both internal engineers and external vendor teams.
What We Offer
- Ownership of the entire third-party integration surface for our AI systems.
- Deep, current work with MCP — the de-facto agent integration substrate — and a real gateway/security stack (OAuth 2.1, OPA, Vault).
- Competitive compensation with a structured performance review process.
- Professional development support — certifications, conferences, and access to emerging tooling.
Collaborative, transparent culture with clear growth pathways toward Staff / Principal engineering.
-
About Softobiz Technologies
Softobiz Technologies is a technology and product services company headquartered in India, operating Global Capability Centers (GCCs) for leading international clients across healthcare, fintech, and enterprise software. Our GCC model enables world-class talent in India to work directly within the product and engineering teams of our global partners, contributing meaningfully to product strategy, growth, and operations.
