AI Engineer with 1+ years in Generative AI & Multimodal Systems
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Final-year Computer Science engineer with hands-on experience building production-oriented Generative AI and Agentic AI systems. Developed a multimodal agentic claims assessment pipeline using Vision-Language Models with a deterministic rules engine and fault-isolated microservice architecture, and a multilingual LLM hallucination detection and remediation framework (CLUE) incorporating RAG-inspired retrieval grounding and multi-stage metacognitive orchestration. Complemented by strong Python and Flask backend experience across RESTful API design, computer vision inference pipelines, and real-time multi-service systems. Actively exploring LangChain, LangGraph, MCP, and LLM evaluation techniques, with an experimental mindset oriented toward rapidly prototyping and deploying AI-driven solutions at scale.
Siksha 'O' Anusandhan University
B.Tech · Computer Science & Engineering
November 1, 2022 – June 1, 2026
Multimodal Evidence Review System
June 1, 2025 – June 1, 2026
Built an end-to-end AI-powered claim assessment platform that emulates insurance adjuster reasoning over cars, laptops, and packages by jointly understanding text and visual evidence. Designed a modular, fault-isolated pipeline - Ingestion → Parsing → Grounding → Vision-Language Analysis → Deterministic Rules Engine → Decision Output - with graceful degradation under provider failures. Implemented structured claim parsing to decompose unstructured conversations into granular damage atoms, supporting multilingual and code-switched inputs and adversarial/prompt-injection resistance. Developed a deterministic rules engine (Supported / Contradicted / Not Enough Information) layered over VLM outputs to reduce hallucinations and produce fully explainable, auditable decisions with risk flags and image-level evidence citations.
CLUE - Cross-Lingual Language Reliability through Uncertainty Estimation [Final Year Research Project]
March 1, 2025 – June 1, 2026
Designed and built an end-to-end hallucination detection and remediation framework for Indian languages (Hindi, Marathi, Gujarati, Odia), integrating multilingual datasets including BHRAM-IL, HaluEval, PsiloQA, and WikiBio into a unified evaluation corpus. Engineered a 5-dimensional reliability feature pipeline - Semantic Similarity, Response Stability, NER Consistency, Numerical Consistency, and Lexical Overlap - with Response Stability emerging as the strongest hallucination signal (Pearson r = -0.40); trained Random Forest, Gradient Boosting, and Logistic Regression classifiers achieving ROC-AUC ~0.85. Implemented a Detect-Inhibit-Rechannelize metacognitive remediation loop: stochastic multi-run generation (3 runs at T = 0.7/0.8/0.9), probabilistic hallucination scoring, and retrieval-grounded response regeneration at lower temperatures (T = 0.3-0.5) with language-keyed fallback templates for Hindi, Odia, and English. Demonstrated language-wise hallucination variation consistent with BHRAM-IL benchmark findings - English < Hindi < Odia - validating the need for language-aware reliability mechanisms in low-resource multilingual AI systems.
SAMVAAD – ISL Translator [SIH’24 Finalist]
March 1, 2024 – September 1, 2024
Designed and developed a real-time RESTful backend using Flask and NodeJS to serve gesture recognition inference to a ReactJS frontend. Built and integrated an NLP + CNN model pipeline achieving 80% accuracy; managed end-to-end API contracts between ML inference layer and frontend services. Led a team of 6 as Full-Stack Developer in HackerWar 5.0 (SIH'24); won among 120 competing projects - demonstrating ownership and delivery under pressure.
CRYONICS – AI Infant Care App
February 1, 2024 – February 1, 2025
Built a Python/Flask backend exposing REST APIs consumed by a React Native mobile client - designed for reliability and low-latency response in a healthcare context. Integrated Firebase (NoSQL) for real-time data storage and user state management; designed data schemas for diagnostic event logging. Implemented a computer vision inference pipeline served via API endpoints, improving caregiver response time by 60%.
AI Engineering
Udemy
June 1, 2026 – Present
Cultural Fit Analysis
The candidate's academic projects showcase a diverse range of applications for AI, from insurance claim assessment to healthcare and sign language translation, indicating adaptability and a broad interest in AI's impact. The focus on multilingual AI systems and low-resource languages suggests an inclusive and globally-aware perspective. The hackathon win as a team lead highlights a collaborative spirit and ability to deliver under pressure, which are positive indicators for cultural fit in a dynamic team environment. However, the candidate is still pursuing a bachelor's degree, which might indicate a need for mentorship and structured guidance in a professional setting.
Soft Skills & Operational Fit
The candidate demonstrates strong problem-solving skills through complex project designs (e.g., fault-isolated pipelines, metacognitive remediation loops). Leadership and teamwork are evident from the SIH Hackathon win as a team lead. The psychometric test score of 382/500 suggests a reasonable work attitude and stress handling, though specific details are limited. The candidate's project descriptions indicate an experimental mindset and a focus on real-world applicability.