Software engineer building reliable backend systems, real-time diagnostic tools, and applied AI workflows.
M.S. in Computer and Information Science, University of Florida
Based in the U.S. and open to relocation, remote, and hybrid roles
I like building software where system behavior matters: backend services that stay observable, control workflows that handle faults predictably, and applied ML systems that can be tested, served, and improved without mystery.
- Distributed systems, backend infrastructure, and service APIs
- Real-time and industrial software with telemetry, state machines, CAN workflows, and fault handling
- Applied AI and machine learning for retrieval, recommendation, ranking, and optimization
- Observability and open-source engineering with Prometheus, Grafana, OpenTelemetry, and CI/CD
| Area | Tools |
|---|---|
| Languages | C++, Go, Python, Java, C#, SQL, Bash, TypeScript, JavaScript |
| Backend and systems | Spring Boot, FastAPI, REST APIs, gRPC, Protocol Buffers, Kafka, PostgreSQL, Redis, RocksDB, OpenSearch |
| Cloud and observability | AWS, Docker, Kubernetes, Terraform, GitHub Actions, Prometheus, Grafana, OpenTelemetry |
| Real-time and industrial | STM32, FreeRTOS, CAN Bus, CANopen, motion control, hardware abstraction, telemetry parsing, diagnostic interfaces |
| Applied AI | PyTorch, semantic search, recommendation systems, neural ranking, CTR prediction, Bayesian optimization, experimentation |
|
End-to-end industrial simulator for STM32-style FreeRTOS device nodes, a C++20 Linux CAN gateway, deterministic CAN fault replay, and a C#/.NET WPF diagnostic workbench. Models initialization, calibration, homing, recipe execution, alarm handling, shutdown, and recovery across virtual CAN devices with replayable JSONL telemetry. Signals: C++, FreeRTOS, CAN, CANopen, WPF, fault injection, diagnostic tooling |
RAG recommender with LLM query expansion, Hugging Face bi-encoder retrieval over FAISS, cross-encoder reranking, and grounded result summaries. Includes a FastAPI serving path, BM25 baseline, and ranking evaluation with nDCG@10, MRR@10, and recall@10. Signals: Python, FastAPI, FAISS, semantic retrieval, reranking, recommendation systems |
|
Click-through-rate prediction system built around a DCN-v2 PyTorch model, deterministic C++ feature hashing, sigmoid calibration, and a production-style gRPC serving API. Supports synthetic local training for the full train, checkpoint, and prediction loop, with Redis-backed calibration overrides for serving. Signals: PyTorch, C++, gRPC, Redis, feature hashing, ranking models |
Gaussian-process Bayesian optimization for a noisy process-yield prediction problem, comparing EI and UCB acquisition functions against a grid-search baseline. Includes process simulation, convergence artifacts, MATLAB equivalents, and a PyTorch regressor trained on simulated noisy measurements. Signals: Python, PyTorch, Gaussian processes, SciPy, scikit-learn, optimization |
|
Hourly .NET batch job that reconciles 24-hour card transaction snapshots into SQLite, records field-level audit history, and tracks ingestion runs. Designed for idempotent snapshot processing with deduplication, revocation, reactivation, finalization, transactional writes, and focused xUnit coverage. Signals: C#, .NET, EF Core, SQLite, batch processing, auditability |
Clear contracts, measurable behavior, small recovery loops, and tests that prove the workflow rather than just the happy path. I am especially interested in backend and infrastructure roles where performance, fault tolerance, observability, and maintainable design all matter. |
I have contributed to Prometheus client_golang, the Go instrumentation library for Prometheus metrics.
- Open pull request:
prometheus/client_golang#2019-[codex] fix metric vec label input aliasing - Focus area: Go metrics instrumentation, API behavior, and regression coverage
For backend, infrastructure, real-time systems, applied AI, or observability-focused roles, reach me at saithej2k3@gmail.com or connect on LinkedIn.