Software Engineer • MS CS @ CU Boulder (’26)
Building high-throughput services, authorization systems, and data platforms.

University of Colorado Boulder | Boulder, CO
Master of Science in Computer Science | GPA: 3.9/4.0 | May 2026

Chaitanya Bharathi Institute of Technology | Hyderabad, India
Bachelor of Engineering, AI & Data Science | GPA: 8.7/10.0 | May 2024

Languages:

Distributed Systems:

• Kafka, gRPC, Protobuf, Redis, REST APIs, OAuth2, Event-Driven Architecture, OpenTelemetry

Cloud & Infrastructure:

• AWS (ECS, Lambda, S3, SQS, DynamoDB), Docker, Kubernetes, Terraform, CI/CD (Jenkins, GitHub Actions)

Databases:

• PostgreSQL, CockroachDB, Neo4j, Redis, MongoDB, BigQuery, DuckDB

AI/ML & Observability:

• LangGraph, LangChain, Ollama, RAG Architectures, Sentence-transformers, Prometheus, Grafana, Jaeger

July 2023 – June 2024 | Chandigarh, India

• Architected an asynchronous fraud detection pipeline on AWS ECS processing 100K+ monthly events with sub-200ms P99 latency, leveraging SQS for buffering and Lambda for event processing, while implementing PgBouncer connection pooling to scale database capacity by 6x (300 → 1,800 connections) and eliminate connection exhaustion during 10x traffic spikes
• Optimized API throughput by 85% (800ms → 120ms) through 3-tier Redis caching with TTL-based invalidation, reducing PostgreSQL load by 70% for 50K+ daily user profile reads and maintaining P99 latency <150ms under peak load
• Reduced deployment time by 87% (2hr → 15min) by automating rollback capabilities in Jenkins CI/CD pipeline with health-check validation and canary deployments, enabling zero-downtime daily production releases

Go, CockroachDB, gRPC, Redis | Sept 2025 – Dec 2025

Engineered Zanzibar-inspired ReBAC system in Go solving multi-tenancy authorization for 10K+ organizations, supporting recursive permission expansion with formal causal consistency proofs validated through Jepsen-style testing.

• Eliminated stale-read anomalies ("new enemy problem") by implementing HLC timestamp-based consistency tokens (Zookies) in gRPC metadata, ensuring linearizable permission checks across distributed CockroachDB clusters
• Achieved sub-10ms P95 latency for direct checks and <25ms for 3-level nested permissions across 100K+ relationships by combining Singleflight deduplication with LRU caching, reducing redundant graph traversals by 80%
• Increased throughput to 10K+ requests/sec using CockroachDB AS OF SYSTEM TIME for stale-consistent follower reads, reducing primary node load by 65% while maintaining strong consistency guarantees

Python, PostgreSQL, LangGraph, FastAPI | Aug 2025 – Dec 2025

Architected zero-cost GraphRAG pipeline integrating local LLM inference (Ollama/Mistral 7B) with 384-dim vector embeddings, achieving sub-200ms semantic search and <50ms graph traversal at depth=3 through LangGraph state machine orchestration.

• Eliminated LLM hallucination by implementing two-phase validation (semantic discovery + deterministic graph verification), achieving 70% node recall and 65% answer relevance by grounding agent reasoning in PostgreSQL recursive CTEs
• Deployed production-grade observability across 4 agent nodes and 6 specialized tools using OpenTelemetry distributed tracing with Jaeger, enabling root-cause analysis of agent decisions with <5ms instrumentation overhead

Go, Kafka, Redis, Docker | Oct 2025 – Nov 2025

Built a distributed microservices platform handling 1,000+ concurrent requests/sec for flash sales with <50ms P99 latency, using Kafka for event streaming and traffic decoupling across order, inventory, and payment services, while implementing sliding-window rate limiting and Prometheus/Grafana observability with real-time alerting.

• Guaranteed strict data consistency by implementing idempotency keys and atomic Redis Lua scripts for inventory management, eliminating race conditions during 1,000 RPS bursts through pessimistic locking and WATCH-based optimistic transactions
• Achieved 99.9% system availability during simulated 30-second payment outages using Circuit Breaker pattern and Dead Letter Queues with exponential backoff, enabling graceful degradation and automatic retry while reducing incident detection time by 60% through anomaly alerting

• Email: yama6766@colorado.edu
• Phone: 303-949-7624
• LinkedIn: linkedin.com/in/yashwanth-mali
• GitHub: github.com/yxshwanth
• Location: Boulder, CO

Building scalable systems, one commit at a time. ⚡