Headline

Senior Software Engineer at Runable | AI Engineer building AI agents and LLM applications | Agent Infrastructure, Sandboxing, Computer Use | TypeScript, Go, Python

About

I build AI agents end to end, from the product people use to the infrastructure the agents run on. At Runable, an AI agent platform with 2M+ users in 218 countries, I own the sandbox infrastructure behind 20k agent chats and 660k tool calls a day. I led its move to a cheaper microVM provider, modeled at 8.3x lower cost with writes about 12x faster, and built the vendor-agnostic layer that lets us swap vendors on live traffic. I also unified the agent runtime so one agent runs in a cloud sandbox or on a user's Mac through a single tool registry. I like building products from zero. On my own I built a computer-use agent that drives a browser and desktop, Runable Flow (a native macOS dictation app, built in 3 to 4 days) and a proactive suggestion engine that ranks what needs a user's attention across every app they connect. On the LLM side I run model evals, engineer context and tool calling, and route model traffic through a server-side gateway. An 8-model eval I ran for dictation cleanup took the graded pass rate from 180 of 219 to 219 of 219. Before Runable I was a Senior Software Engineer at Pixxel, where I built the order desk for hyperspectral satellite imagery, Go backend services, and an adapter architecture that made LLM features pluggable and cut development time 55%. At SixSense I made an AI defect analytics product 4x faster, which helped win major clients and lifted conversion 35%. I work in TypeScript, Go, Python, Rust and Swift, on Linux, microVMs, Docker, Kubernetes, Google Cloud and Cloudflare Workers.

Location

Bengaluru

AI stack

claude code

github copilot

E2B

Kimi Code

Retell AI

Zed

Work experience

Senior Software Engineer at Runable, Jun 2026 - Present · 5 mos, Bengaluru, Karnataka, India, AI agent platform with 2M+ users in 218 countries. Merged 415 PRs in about 14 weeks and reviewed 503. Agent infrastructure • Owned the sandbox infrastructure behind 20k agent chats and 660k tool calls a day, and led its move to a cheaper microVM provider, modeled at 8.3x lower cost with writes about 12x faster. • Built a vendor-agnostic sandbox layer to swap vendors on live traffic with no schema migration across 963k chats and a rollback-safe workspace migration. • Unified the agent runtime so one agent runs in a cloud sandbox or on a user's Mac, through one tool registry and a CI parity gate, closing a 22 vs 5 tool gap and deleting about 9k lines. AI products built from zero • Built a computer-use agent solo that drives a browser and desktop. • Built Runable Flow, a native macOS dictation app, solo in 3 to 4 days, with Swift 6, pinned code signatures, Keychain-backed device enrollment, auto-updates and deploy-safe usage billing. • Ran an 8-model eval for dictation cleanup that took the graded pass rate from 180/219 to 219/219, and built a server-side model proxy so provider keys and billing never sit on a user's device. • Built a proactive suggestion engine in 2 weeks that scans every app a user connects and ranks what needs their attention by priority and deadline, with a one-click agent action for each. Production, security and team • Owned reliability and security platform-wide, from a chat streaming rebuild that ended errors hitting about 3,000 users in 2 weeks to website previews, deployments, auth, SSRF and XSS. • Built monitoring and paging from zero with CI gates so no service ships unmonitored, and an OpenTelemetry pipeline routing all telemetry through one gateway. • Rebuilt the marketing site with per-route server rendering for a Lighthouse SEO score of 100, made slide-deck renders about 5x faster, and cut mobile page images 91%. • Owned hiring alongside engineering and closed about 10 roles, several of them senior, in under a month.

Senior Software Engineer at PIXXEL, Jun 2024 - Jun 2026 · 2 yrs 1 mo, Bengaluru, Karnataka, India, Hyperspectral satellite imaging. • Built the platform's order desk from scratch, where customers request, buy and view hyperspectral satellite imagery, and built Go backend services alongside the frontend. • Built AI route automation for 8 routes, saving about 4 weeks of effort each at near-100% accuracy. • Designed an adapter architecture that made LLM features pluggable across the product, cutting development time 55%, and drove team-wide adoption through code walkthroughs. • Built a zero-dependency reactive state manager that improved rendering performance 40%. • Led the migration of a 4-year-old codebase to 65% TypeScript in 10 months. • Merged sprint, QA and release into one cross-team flow with under 10% release overflow. • Built an accessible (WCAG) component library from Figma with designers, making UI development 30% faster. • Standardized external package usage, cutting replacement time 95%.

Frontend Engineer at SixSense, Jul 2022 - May 2024 · 1 yr 11 mos, Bengaluru, Karnataka, India, AI defect analytics for semiconductor manufacturing. • Made the product 4x faster to load and render by restructuring its architecture and upgrading key libraries, which helped win major clients, lifted conversion 35% and was one of the headline results in the year-end company presentation. • Led an R&D restructure that cut bugs 85% and development time 70%. • Built an image upload system handling 200,000+ images at 60 fps and a canvas annotation tool rendering 2M+ points at 60 fps. • Automated per-customer CI/CD builds with GitHub Actions, cutting deployment failures 95%. • Moved auth to server-side cookies and cut average API response size 89% with compression. • Mentored a team of 3 developers and oversaw frontend hiring.

Frontend Developer at NowPurchase, Dec 2021 - May 2022 · 6 mos, Kolkata, West Bengal, India, - Implemented tailored technologies for the non-technical metal industry, enhancing real-time organizational visibility with an 80% reduction in cross-team communication delays. - Experimented with different project ideas involving AR, OS listeners, and multi-network connections, for extending product scope, enriching features, and improving product pitches. - Played a key role in establishing connections within the industry, which helped us in better visibility of the problems faced on the ground on a day-to-day basis. - Conceptualised and executed many MVPs within record time, resulting in some big client onboarding and an increased client conversion rate of 30%.

Frontend Developer at GreedyGame, Aug 2021 - Nov 2021 · 4 mos, Bengaluru, Karnataka, India, - Enhanced business site performance by 40%, achieving a Lighthouse score of more than 95%. - Integrated analytics and error monitoring systems for efficient bug reporting and user analytics. - Worked intensively with visualization tools and charts to provide users with proper insights about the data. - Created and maintained in-house UI component libraries, reducing development time by 55%.

Frontend Developer at GreedyGame, May 2021 - Aug 2021 · 4 mos, Bengaluru, Karnataka, India, - Worked on stabilising the products by debugging the existing bugs. - Worked on improving the performance of existing features and help ship new features.

Education

Bachelor of Technology - BTech, Information Technology, Kalinga Institute of Industrial Technology, Bhubaneswar

Public links

github: https://github.com/singh-pk