Building backend systems, distributed infrastructure, and applied ML products with a focus on reliability, performance, and clean execution.
“We used to look up at the sky and wonder at our place in the stars.”
I'm Dhruv, an undergrad studying Information Science at B.M.S. College of Engineering in Bangalore (graduating 2027).
Most of my time goes into building things, reading, and falling down rabbit holes trying to understand how systems actually work under the hood. I enjoy the quiet craft of figuring things out from first principles, writing clean software, and tinkering with ideas until they work.
SDE Intern — Profinch Solutions (July 2026 — Sept 2026)
Worked on backend systems with a focus on thoughtful implementation, performance, and clean execution.
B.M.S. College of Engineering, Bangalore (2023 — 2027)
B.E., Information Science & Engineering
Malicious URL detection via probabilistic data structures. Sized a 5.14 MB Bloom Filter for 3M URLs at 0.1% target false-positive rate, indexing 268,970+ threat URLs into 1/100th the memory of a hash set. Two-stage lookup pipeline handles fast-path screening at ~0.003ms with 145,000 URLs/sec throughput, querying PostgreSQL only on hits. Validated 0 false negatives and 0 false positives across 100k test URLs. Published on PyPI.
Natural language to SQL platform enabling instant English-to-SQL analytics over CSV, Excel, and JSON datasets. Two-stage RAG pipeline retrieves DDL embeddings and curated SQL examples from ChromaDB to ground Gemini 2.5 Flash SQL generation, executed against isolated per-user PostgreSQL schemas. Event-driven architecture with Kafka across 4 topics and Redis sliding-window rate limiting.
Distributed rate limiting engine supporting 4 algorithms (sliding window log, token bucket, leaky bucket, fixed window) with atomic Redis Lua script execution. Load tested to 4,000+ RPS at ~12ms p50 latency with hot configuration reloading without service downtime, WebSocket metric streaming, and PostgreSQL persistent audit logging.
AIS vessel anomaly detection pipeline flagging suspicious ship maneuvers, transponder manipulation, and AIS spoofing. Processed 7.3M vessel pings and scored 4,621 ships in ~3 minutes, generating interactive geospatial anomaly maps and trajectory reconstructions.
Graph Attention Network for social bot and Sybil account detection. Analyzed 14,368 accounts, achieving 97.12% classification accuracy with statistically validated attention behavior (p < 10⁻⁵). Integrated GNNExplainer to provide interpretable subgraph explanations for flagged nodes.
Machine learning pipeline retrieving 5 gas abundances and equilibrium temperature from exoplanet transmission spectra (Ariel Big Data Challenge dataset). Tuned XGBoost models generalized best to real JWST NIRSpec observations of WASP-39b, matching published retrieval results on out-of-distribution real observational data. Research manuscript under review at Springer Nature.
Physics-informed 1D convolutional neural network for radio modulation classification from raw IQ signal data across 11 modulation types. Integrated CBAM spatial and channel attention to achieve 93.06% high-SNR accuracy with only 1.26M parameters.
Fully offline, privacy-first PDF question-answering application using Retrieval-Augmented Generation. Embeds documents into ChromaDB and queries local LLMs via Ollama with automated evaluation tooling.
Real-time network intrusion detection system pairing a Bidirectional LSTM neural network with live packet capture via tshark. Achieved 99.16% detection accuracy and 0.83% false-positive rate across live packet streams.
Lightweight terminal clipboard history manager supporting text and image capture, fuzzy search, and persistent SQLite storage. Published on PyPI.
C / C++ / Python / Java / JavaScript / FastAPI / Express.js / React / TensorFlow / Redis / PostgreSQL / MySQL / MongoDB / Docker / Kubernetes / AWS / Git / GitHub / CI/CD (Render) / Postman / Linux / Shell Scripting / Data Structures & Algorithms / Operating Systems / DBMS / Computer Networks / OOP / Machine Learning / Distributed Systems