Rohit Rajan

Lead a 14-person engineering team across backend and frontend. Hands-on senior IC who writes production code daily, runs code review for backend PRs, and owns architecture across our consumer products. Day-to-day is roughly 50% coding, 50% design, review, and people leadership.

Applied AI and Agentic Systems

• Built a timestamp-aware RAG pipeline over an internal corpus using LangChain tool wrappers. The pipeline returns the specific passage plus surrounding context instead of whole documents. I added evaluation scripts that compare retrieval strategies against a held-out question set, so prompt or retriever changes are graded by data rather than feel.
• Designed an embedding service using Gemini 768-dimensional embeddings as a separate FastAPI worker that batches and retries independently from ingest. Transient upstream failures (model API timeouts, rate limits) do not block uploads, and throughput scales with worker count rather than request rate.
• Migrated production vector search from FAISS to Qdrant so metadata filtering happens inside the HNSW traversal instead of as post-filtering. Kept recall high for users with large libraries, where the earlier FAISS-with-post-filtering setup was returning near-empty result pages.
• Designed structured logging and basic telemetry around LLM calls, including token counts, latency percentiles, tool-call traces, and failure reasons. This is the layer that turns prompt or model regressions into a data problem instead of a user-complaint problem. AI-Assisted Development Practice
• Use Claude Code and Cursor as my daily editors across Python and TypeScript work. Built a personal rhythm of scoping changes with the agent, letting it draft, then reviewing diff-by-diff. The team has adopted the same workflow, which has tightened our review cycle on routine refactors and test-writing.
• Evaluate, validate, and harden AI-generated code through code review and test coverage gates before merging. Wrote internal guidance on what we always check in AI-assisted PRs (correctness, tests, observability) and what we deliberately do not block on (style, taste).
• Continuously experiment with emerging AI tools across the team. Recent work includes evaluating MCP-style tool wrappers for internal automation and prompt-engineering patterns for repetitive engineering tasks. Backend Engineering and Distributed Systems
• Decomposed a legacy Node.js monolith into domain-owned services. Wrote the migration playbook, ran a tenant-by-tenant cutover with reconciliation gates, and kept the legacy path live as source of truth until the new path passed two full business cycles.
• Built a Python personalized feed service using a fan-out architecture, caching processed feeds in Redis to hold sub-50ms response times under production load.
• Debugged a production outage where a feed service was timing out only for large accounts. Traced it to unbounded collection queries and N+1 lookups, redesigned the read path to use paginated queries with database-level indexes, and brought tail latency back under SLA the same day. Documented the anti-pattern and found two more services with the same issue before they became incidents.
• Migrated production workloads from AWS (EC2, S3) to a hybrid setup with self-hosted Hetzner bare-metal fronted by Nginx for compute. Cut infrastructure cost by 60 to 70 percent while giving us more headroom per service. Cross-Functional Partnership and Mentorship
• Partner closely with product and design on roadmap and UX direction. Translate ambiguous product asks into engineering briefs with concrete options, latency targets, and recommendations that non-engineers can act on.
• Mentor engineers across junior, mid-level, and senior levels through weekly 1:1s, code-review pairing, and progression conversations. Run end-to-end hiring for engineering roles, from sourcing through onboarding.
• Investigated the axios npm supply-chain incident against our deployed dependency graph, confirmed none of our servers were compromised, and formalized a lightweight dependency review step for future releases.

Built custom e-commerce platforms in PHP and Laravel with end-to-end payment gateway integration and inventory management.
Wrote hand-tuned SQL for sales reporting where ORM-generated queries were too slow, keeping client dashboards responsive at peak query times.

Developed core modules for a custom CRM, handling data flow for sales leads and operational tracking.
Designed and documented REST APIs consumed in production by partner mobile applications and external integrations.

Maintained Windows Forms desktop applications, performed routine database maintenance, and wrote stored procedures for reporting.