Hello,
I'm
Software Engineer — Backend & Infrastructure
IIIT Gwalior '26 · CNCF Contributor · IEEE Published
More About Me
Hello,
I'm
-> Sole owner of a rosbag replay & regression-gate framework: replays real robot field data through any code version of any of 6 autonomy modules in isolation and blocks regressions in CI before they reach the robot.
-> Built an always-on C++ flight recorder — MCAP ring buffer with O(1) hardlink incident preservation, validated at 28.65 MB/s sustained in 71 MB RAM on robot hardware.
-> Built the fleet observability pipeline (rsyslog/RELP → Vector → Loki/Prometheus → Grafana) and freed 70% of NVIDIA Orin compute via the V2 backend migration.
-> Independently shipped 10+ integrations (LiveKit, Pipecat, LangChain, CrewAI, LlamaIndex), a production MCP server, and an n8n community node published on npm.
-> Built a WhatsApp TTS data-collection platform enabling speech-dataset creation for the low-resource Maleku language; cut microservice CI build time by 67%.
-> Engineered the OpenTelemetry metrics adapter for Jaeger V2, improving metrics-collection speed by 40% and establishing observability parity between V1 and V2.
-> Streamlined the OTEL Collector (TracerProvider/MeterProvider) and integrated the health-check extension across V2 components.
-> Developed a multi-level user onboarding, match allocation, and report generation system serving 10,000+ registered users.
-> Built reusable React.js components and a Node.js + Express.js + MongoDB backend exchanging data via REST APIs.
-> Built NLP-based candidate-assessment pipelines using sentiment analysis (VADER, Cohere) and a noise-suppression model improving transcription accuracy by 30%.
-> Led unit and widget testing for production Flutter apps, reaching 95% test coverage, and built QA test suites for MERN websites and Flutter apps.
view project
DataLens — Agentic Data Catalog Assistant
A context-aware agentic assistant for an enterprise data catalog: answers natural-language questions with grounded, column-level citations and a confidence score on every answer. A deterministic governance guard hard-blocks unsafe queries before any LLM sees them, and a critic loop enforces groundedness with bounded retries (LangGraph guard → planner → executor → synthesizer → critic). Hybrid retrieval via ChromaDB + BM25 with reciprocal-rank fusion; generated SQL executes against a real store.
view project
Deep-Learning Intrusion Detection (IEEE Publication)
Peer-reviewed research published on IEEE Xplore: a network intrusion detection system trained on the CIC-IDS dataset (100,000+ samples), using DSSTE sampling to correct severe class imbalance and reaching 99.5% accuracy on threat classification.
view project
Order-Management gRPC Microservice
A Go microservice for order management built on gRPC and Protocol Buffers — typed service contracts with clean separation between transport and business logic, demonstrating production microservice patterns in a compiled, statically-typed stack.
view project
Sentiment Palate
An NLP service that classifies user reviews in real time with 95% accuracy. Containerized with Docker and deployed on Kubernetes for orchestration and load balancing, with a FastAPI backend and a React frontend.
Six internships including Origin (robotics) and Camb AI (speech AI) — shipping production systems with real users, real hardware, and real deadlines.
Code merged into Jaeger (CNCF) through maintainer review as an LFX mentee — building the OpenTelemetry metrics adapter for Jaeger V2.
Peer-reviewed research on deep-learning intrusion detection published on IEEE Xplore — solving class imbalance with DSSTE sampling.
C++, Python, and Go across Docker/Kubernetes, CI/CD, and observability stacks — performance measured on real hardware, never assumed.
+91 7406320038
saransh.shankar@gmail.com
Bengaluru, India
Category - Web Application
Lorem ipsum dolor, sit Lorem ipsum dolor sit amet consectetur adipisicing elit. Quidem perferendis corporis culpa! amet consectetur adipisicing elit. Hic doloribus labore assumenda nostrum commodi. Officia. Lorem ipsum dolor sit amet consectetur adipisicing elit. Incidunt.
