// Backend & ML Systems Engineer
K Ashraf Ahmed
Contributor to Zalando Skipper, CNCF etcd & CALL-E — 8 merged PRs running in production, including a load-balancing algorithm.
Associate Software Engineer @ The Jobs Jungle · Bengaluru, India · open to backend & ML-infrastructure roles
weightedRoundRobin Load-Balancing Algorithm
Designed and shipped a smooth weighted round-robin algorithm with dynamic health-derived endpoint weights. Defended the locking design under maintainer review with a self-built C benchmark harness sustaining ~31k req/s at 1,000 endpoints — the harness was kept in the repo for testing all algorithms.
lifoWithBody / lifoGroupWithBody — Concurrency-Limit Fix
lifo released its concurrency queue slot as soon as the backend sent response headers, so a backend streaming a long body left the slot free and the route admitted past MaxConcurrency. Added lifoWithBody and lifoGroupWithBody that hold the slot until the body finishes streaming, using a separate state-bag key so a plain lifo() on the same route can't pop the release function early. The regression test fails against the old filters — they wrongly return 200 where the new ones correctly return 502 on queue timeout.
mobilize — Wave Dispatch Agent + MCP Server
Shipped an agent app and MCP server that treats a model's stated confirmation as a noisy signal rather than a fact — scoring call evidence against historical show-rate before it counts. A 300-trial evaluation harness against a synthetic population with known ground truth measured 94.6% accuracy for the calibrated policy vs 87.7% for trusting every stated yes. Crash-safe dispatch via a write-ahead ledger, verified with a real SIGKILL'd subprocess. 89 tests.
Prometheus Native Histogram Support
Implemented opt-in native histogram support with OTEL-recommended bucket tuning, deploying to production Kubernetes clusters while maintaining backward compatibility with existing dashboards.
Per-Data-Client Load Latency Metrics
Added instrumentation to measure route-loading latency per data source, giving SRE teams visibility into which integrations are slow.
TLS Certificate Matching Documentation
Clarified security documentation for certificate-matching semantics across v3.5–v3.7, resolving a 14-month-old confusion issue about wildcard support. Merged by a core maintainer.
Go Toolchain Security Updates
Patched three CVEs by bumping Go to 1.25.11 as part of etcd organization's tracked security effort.
Associate Software Engineer — The Jobs Jungle
Bengaluru, India · Aug 2026 – Present
Develop and maintain REST APIs for the applicant-tracking platform on Java 17 / Spring Boot 3 over MongoDB, securing endpoints with JWT + OAuth2 (Google, LinkedIn) and role-based access control. Write JUnit and Spring Security tests against OpenAPI-documented service contracts; triage and resolve issues across development, staging, and production environments.
Aegis — Predictive Observability Platform
Automated microservice failure prediction and root-cause analysis: an event-time Kafka stream correlator whose topological RCA was correct on all 78 evaluated incidents, a statistical detector (EWMA z-score + OLS trend projection) reaching 228-second median lead time, and a HistGradientBoosting classifier evaluated in shadow mode against it with PSI drift monitoring, a versioned registry and gated retraining.
Go + Java Auth Platform — Polyglot Microservices Identity System
Built a two-service identity platform: a Go auth service with rotating refresh tokens, reuse detection, and a Redis JTI blacklist, consumed by a Java/Spring Boot resource API that validates every request over gRPC — with a token-hash Caffeine validation cache and a fail-closed Resilience4j circuit breaker. 61 tests (Testcontainers + in-process gRPC) in CI; k6-load-tested at ~92 req/s with 6ms p50 across the two-service auth path.
Custom Key-Value Storage Engine
Built a durable, crash-resilient storage engine from scratch in Python, achieving ~26,000 writes/sec in batched fsync mode and 100% recovery rates under simulated unclean shutdowns (kill -9), by designing an LSM-tree architecture with a Write-Ahead Log, Bloom filters, and size-tiered compaction.
ClearText API — Async ML Inference Platform
Optimized toxic comment ML inference throughput, scaling to 231 requests/sec at 106ms average latency under a 500 concurrent-user load, by building an asynchronous FastAPI serving backend utilizing Celery task queues, Redis model-versioned caching, and a custom worker micro-batcher.
Distributed Log Processing & Analytics System
High-throughput real-time log processing platform ingesting 343 logs/sec with 10ms P50 latency and 99.5% processing reliability, with real-time alerting and analytics.
Radar–Camera Fusion for Real-Time Drone Detection (YOLOv8n + LSTM)
Confidence-weighted late fusion of YOLOv8n detection with radar-inspired motion features, lifting precision from 0.91 to 0.94 (F1 0.92) while holding 8.3 FPS real-time inference on CPU — no GPU required at serving time. An LSTM trajectory forecaster cut RMSE from 1.78 to 1.64 versus a GRU baseline. Two peer-reviewed IEEE papers (ASIACONF 2026, C2I6 2025).
Skills & Tech Stack
About
Backend and ML-systems engineer in Bengaluru. I build infrastructure that behaves like production rather than like a demo: a storage engine with crash-recovery proofs under kill -9, an inference platform with circuit breakers and a dead-letter queue, and an ML pipeline whose retraining is gated on offline evaluation. Currently an Associate Software Engineer at The Jobs Jungle, working in Java/Spring Boot, and a contributor to Zalando Skipper, CNCF etcd and CALL-E.
New Horizon College of Engineering, Bengaluru
B.E in Computer Science and Engineering | GPA: 7.71 | Nov 2022 – July 2026
Coursework: Data Structures & Algorithms, Design & Analysis of Algorithms, Operating Systems, Computer Networks, Database Management Systems, Linux System Programming, Computer Architecture (ARM), Machine Learning, Generative AI, Cloud Architecture & Security
Publications
IEEE ASIACONF 2026 — "Lightweight Radar-Camera Fusion for Real-Time Drone Detection and Trajectory Prediction" (peer-reviewed): confidence-weighted late fusion of YOLOv8n with radar-inspired features lifted precision 0.91 → 0.94 at 8.3 FPS CPU-only inference.
IEEE C2I6 2025 — "Lightweight Real-Time Drone Detection and Trajectory Forecasting Using YOLOv8 and LSTM Networks" (peer-reviewed): mAP 0.89 at ~9 FPS CPU-only inference.
Get in Touch
Interested in working together or have a question? I'd love to hear from you.
ashrafahmed1232@gmail.com