Priyansh Kansara

Applied ML & Embedded Systems

2nd-year Electronics & Telecommunication @ KJ Somaiya College of Engineering · CGPA 9.15/10 · Mumbai, India

58.6K lines of Python 909 tests 3 merged OSS PRs 9 PRs under review

Projects

Vishwas

Zero-cloud verification platform · ~34K LoC · 445 tests

Users message a number on WhatsApp; URLs, executables/APKs, documents (OCR) and images/audio get checked and answered with a plain-language verdict plus a confidence band, in 7 languages. Runs CPU-only on an i5 laptop — no cloud calls. Routing decisions are deterministic; an LLM only narrates already-finished verdicts, behind a prompt-injection guard.

UARTScope

"Wireshark for microcontrollers" · ~13.9K LoC · 160 tests

Bidirectional UART/I²C/SPI/CAN/Modbus decoders, CAN from vendor DBC files, live charts, threshold alerts, session record/replay with golden-baseline diffing, a plugin marketplace, and an MQTT bridge to existing dashboards. Pure Python, ~50 REST/WebSocket routes, GitHub Actions CI.

CiteSure

Local citation verifier · ~9.7K LoC · 288 tests

Checks whether a cited source actually supports a claim, not just whether the link is alive — three tiers from deterministic link/status checks through content extraction to an opt-in NLI cross-encoder. Flags dead, retracted and paywalled sources. Abstention is a first-class output: unverifiable claims return "can't verify" instead of a hallucinated verdict.

blind-mcp

Company-research MCP server · ~1K LoC · 16 tests

An MCP server for researching companies on Blind, built on an upstream MIT scaffold and extended: ranked thread search against a natural-language question, with employer-tag and date provenance preserved on anonymous posts.

Open source

3 merged PRs into PyPA-ecosystem tooling:

9 more open and under review at pytest, pytest-xdist, pandas, Hugging Face diffusers, the MCP reference servers, LMCache and Axl.

Experience & education

AI Product Intern — ACSE (Jun–Jul 2025)
  • Replaced a company chatbot's keyword-search routing with LLM response classification on a small Llama foundation model, evaluated against real user queries.
  • Added OCR + metadata extraction for document inputs; iterated on response accuracy.

B.Tech, Electronics & Telecommunication — KJ Somaiya College of Engineering, 2025–2029, CGPA 9.15/10

IBCP Diploma in Artificial Intelligence — Elpro International School, 2023–2025 (capstone: waste-type classification model for automated recycling)

LinkedIn GitHub Email