Job Title:  Consultant | Backend Developer - Java | Delhi | Operations, Industry & Domain Solutions | DigiGov

Skills
Languages C, C++, Python, Java, SQL, Bash
Systems Linux, POSIX APIs, Shared Memory IPC, Process Scheduling, Multithreading, Virtual Memory,
TCP Networking
Architecture gem5, RISC-V, Computer Architecture, Memory Hierarchy, Cache Behavior, Instruction Set
Architectures
Compilers Lexical Analysis, Recursive-Descent Parsing, Symbol Tables, IR Generation, Assembly-like
Code Generation
ML / Tools Deep Reinforcement Learning, DQN,

Developed SafeTail, a deep reinforcement learning scheduler for minimizing tail latency in heterogeneous
CPU–GPU edge-computing environments.
• Modeled scheduling as an MDP with a 31-action server-subset space across 5 edge servers,
enabling redundancy-aware request placement and first-response selection.
• Implemented an integrated Encoder + DQN architecture that maps variable-length CPU, GPU,
memory, and network-state vectors into a fixed 24-dimensional embedding.
• Designed a QoS-driven reward function combining resource availability, deadline satisfaction,
and M/M/1 queue waiting-time penalties.
• Built task-specific MLP latency predictors from real hardware traces and evaluated on 500,000
requests across object detection, image segmentation, and audio workloads.
• Benchmarked against MinLoad, MinProp, and Random policies across 50th, 90th, 95th, and
99th latency percentiles.

Education:
Btech