SRIKANTH KADABA

• Coding assistants Agents that edit real repositories
• Sandboxed executions Workflows that run in a contained runtime
• Eval harnesses Score the work before it ships
• Speech to text Voice into structured intent
• Text to speech Spoken answers from the system

Across the stack — coding, sandboxed runs, evals, speech in, speech out.

• OR solvers Math for routing and scheduling
• Attention-based VRPs Learned attention on vehicle routes
• Optimizations Better plans under real constraints
• Unsupervised clustering Finding structure in geo demand
• Geospatial grids Cells for dispatch and coverage
• ALNS Neighborhood search for better routes
• Last-mile delivery From demand to the stop
• Time series forecasting Demand as it changes over time
• Discrete event simulation Simulate the system before rollout

Routing, capacity, demand, and simulation.

• HPC Cluster jobs for heavy simulation
• Inference Models answering in production
• Slurm Batch work on the cluster
• Go servers Backend services in Go
• High scale Lots of rows, lots of locations
• gRPC Calls between services
• Android full-stack App through to the backend

HPC, serving, and the systems in between.

• Go GC

  Ask me how I know the collector doesn’t run on a cron.

• gRPC sidecars

  Ask me how sideloaded multiplex cut model latency.

• Redis cache

  Ask me how I freed up 14GB of string key-value pairs, with binary blobs instead.