Resume
B.Tech. in Computer Science & AI · Plaksha University
SDE ResumeBackend, full-stack & fintech-focused
AI/ML ResumeResearch, RAG pipelines & LLM systems
Contact
Education
Plaksha University
B.Tech. in Computer Science & AI
CGPA: 8.20
Aug. 2022 – July 2026
Skills
Languages
PythonSQLJavaScriptTypeScript
AI / ML
PyTorchLangChainOpenCVRAG PipelinesSegment Anything (SAM)
Web & Backend
FastAPINode.jsNext.js (SSR/CSR)REST APIsPydanticTailwind CSS
Cloud & Tools
Azure (Cosmos DB, OpenAI)FirebaseFAISSGit/GitHubVercel
Research Experience
Undergraduate Researcher
Visual Benchmarking of VLMs | Sup. Prof. Pankaj Pansari · Plaksha University
- Constructed a 459-image benchmark with instrument-verified ground truth across 5 physical estimation tasks; personally handling all instrument measurements and building a 3-tier blur-degradation pipeline (OpenCV) to test model perceptual robustness.
- Benchmarked 6 VLMs (Gemini 3.1 Pro, GPT-5.4, Gemini-Robotics-ER, Claude Opus 4.7, Gemma 4, Qwen 3.5) across 3 independent runs each, finding performance collapses significantly under image blur across all frontier models.
- Identified a misidentification-vs-miscalibration failure dichotomy via top-10 error analysis per category: GPT-5.4 fails by misidentifying objects, while Gemini variants correctly identify but overestimate weights by 1.8–2.5.
Undergraduate Researcher
Supervised by Prof. Sandeep Manjanna · Plaksha University
- Developed preprocessing pipeline (contrast adjustment, noise reduction, edge detection) to optimise Segment Anything Model (SAM) for agricultural crop-weed segmentation on sparse datasets, maximising zero-shot segmentation quality on out-of-distribution agricultural imagery.
- Improved batch processing throughput for 10,000+ image datasets by implementing vectorized NumPy operations and Python multiprocessing.
Industry Experience
Freelance Web Developer
PQRS Research (Dr. Niket Tandon) · Kolkata
- Leading a full redesign and feature buildout of the PQRS Research website in Next.js, including a dual onboarding system with distinct registration and intake flows for mentors and mentees.
AI Engineering Intern
Cotality ·
- Architected a FastAPI backend with AST-based parsing for 15+ languages, enabling automated documentation generation across polyglot codebases at 50,000+ LOC/day.
- Cut AI embedding costs by 85% by building a hash-based system that fingerprints each function's AST and skips re-embedding unchanged code, eliminating redundant Azure OpenAI API calls.
- Built a RAG pipeline using LangChain and Azure OpenAI, backed by Cosmos DB with IVF vector indexing for semantic code retrieval –- migrated from FAISS to support persistent, production-scale storage.
- Shipped 12 production REST API endpoints with typed Pydantic schemas, powering real-time documentation preview in the client-facing web interface.
Freelance Web Developer
Anuj Desai Associates ·
- Delivered a full-stack site with a public careers page and a Firebase-backed admin dashboard, giving firm partners a private portal to review and manage job applications.
- Secured admin access via Google OAuth with an email allowlist (Firebase Auth), enforcing partner-only access without a separate auth backend.
- Built a GitHub-editable content layer so the client can update site copy post-handoff without developer involvement.
Software Development Engineer Intern
Orangewood Labs (Robotics) ·
- Designed modular LLM task-planning system for a food-preparation robotic arm, decoupling high-level intent logic from low-level actuation to enable parallel development across firmware and AI teams.
Leadership
Athleda – Plaksha Sports Society
Head of Technology · Plaksha University
- Built and maintained a sports platform serving 500+ students, shipping event registration, live-scoring, and photo galleries, while mentoring 2 junior developers through code reviews and production deployments.
LEAP.AI @ Plaksha
Web Developer → Head of Web Development · Plaksha University
- Contributed as a core developer on a team of 7–8, then led the team through 2 full platform redesigns as Head, growing the AI club site to 800+ monthly users for events, workshops, and project showcases.