AI Systems Engineer · Chennai, India · Remote
I design and build AI systems, from the first idea to production.
Agents, agentic workflows, retrieval and memory, plus the backends, data models and infrastructure underneath them. Shipping software since 2020; today a senior engineer on a manufacturing SaaS platform, and building my own agent tooling on the side.
- Your time
- --:--:--
- Mine, Chennai
- --:--
- Problem breakdown · 2026-11-09
- Item-not-received chargebacks are lost to silence, not to fraud
Add me to your agent
claude mcp add --transport http devjags https://thedevjags.com/mcpHow did Jags stop retried agent runs from writing twice?
read_work switchboard · Each write is keyed by its run step, sha256(runId, action, callIndex), not by its input, so a retry that rephrases an issue title still writes once. Switchboard
What I build
Agentic systems
Multi-agent orchestration, tool and MCP design, working memory, and the review and verification loops that make agent output trustworthy.
AI products, 0 to 1
From a vague problem to a shipped product: framing, data model, APIs, interface, deployment, and the feedback loop after launch.
Foundations
Kotlin and Spring Boot services on PostgreSQL, six ERP and accounting integrations, and infrastructure as code on AWS and GCP.
Selected systems
All work →- Agentic Workflow SystemIn daily useOne kernel file, a stage ladder, file-based working memory, and enforcement hooks shared by Claude Code, Codex, OpenCode and Cursor.
- Tachyonv1.1.1 releasedSingle-binary code reader and review workbench: live diffs across worktrees, pinned intent cards, Rust and Go engines.
- SwitchboardSandbox deployConnectors for Intercom, Slack, Linear and Attio; agents declare what they may do; a ledger gates every write.
- Ad StudioIn internal useGuarded browser capture, agent-written scripts with human approval, HTML-to-video rendering, local voice, and loudness QA.
- Pulse and AgentLensIn personal usePulse, a contribution dashboard with a desktop app, then AgentLens, a focused rewrite for agent sessions.
- SuttruLiveVoice translation, a sign and menu reader, grounded answers, and transit routing in an offline-capable PWA.
Daily
All notes →One short note every morning: a real problem broken down, a field note on AI engineering, or one concept explained.
- 2026-11-09No. 025Problem breakdownItem-not-received chargebacks are lost to silence, not to fraudSmall sellers lose many card disputes they could have won because the evidence is scattered across four tools and the deadline passes first. The fix is mostly plumbing, with a narrow and well-fenced job for a model.
- 2026-11-08No. 024ConceptAgent memory that only appends will remember both answersLong-term memory for agents is usually a vector store that keeps growing, and similarity search has no idea which of two contradictory facts is still true. The fix is to treat memories as dated claims that newer claims can retire.
- 2026-11-07No. 023Field noteAgent SDK hooks can now fail closed, so make it your defaultClaude Code 2.1.317 lets a failing Agent SDK hook block the action it guards instead of waving it through. A guard that disappears when it crashes was never a guard, and this changes how policy hooks should be written and tested.
Writing
All writing →Working together
I am open to AI engineering roles and a small number of zero-to-one builds where the hard part is the system, not the screen. The fastest way to reach me is email.