Amazon

• Building Unified Compliance Intelligence for Amazon Marketplace, focused on detecting non-compliant listings and turning compliance signals into clear seller-facing guidance.
• Building backend services with Java, Coral, ECS Fargate, and Amazon Bedrock to help flag products, explain failed requirements, and surface the documentation updates sellers need to make.
• Using Claude/Sonnet workflows to translate policy and product signals into plain-English compliance explanations while keeping service boundaries auditable.

Apple

• Built production tooling for macOS and telecom engineering workflows across LTE/NR message validation, protocol debugging, and internal developer support.
• Architected an LLM-powered triage automation pipeline using retrieval over historical logs, CI/CD integrations, and incident clustering to help engineers reason over failures faster.
• Led end-to-end bring-up of a cellular testing lab in India using iPhone and USRP X410/B210 setups, with Python automation for repeatable setup and testing.

PayPal

• Improved Global Merchant Lending admin tooling and audit workflows with React, Node.js, and internal PayPal services for compliance-sensitive financial operations.
• Reduced manual review friction for operational teams by improving data visibility, state handling, and lending-review controls.

Neatlogs

• Hardened TypeScript SDK transport behavior to prevent duplicate API writes after network failures, validating 5,350 fault-injected mutations with zero replays and packaged consumers across Node.js 18-24.
• Performed end-to-end qualification across Python and TypeScript SDKs, the instrumentation Wizard, standalone Skills, and setup documentation, validating 129/129 development trace exports and exact persisted dashboard traces.
• Built deterministic test harnesses covering streaming, tool calls, workflow hierarchy, errors, cancellation, privacy and masking, concurrency, lifecycle management, and endpoint isolation.
• Developed cross-language observability tooling and prototypes for Trace Doctor diagnostics, provider reliability, prompt workflows, migration Skills, and cost and trace-replay analysis across Python, TypeScript, and Go.

Apple

• Created an internal single-page application (SPA) using React.js, Tailwind CSS, Spring Boot, Aluminum, Splunk, and internal Apple frameworks, achieving an 83% reduction in support turnaround time by optimizing cross-domain access and reducing workflow complexity.
• Designed and deployed an ML pipeline (XGBoost + feature engineering) to predict SLA resolution times with 80%+ accuracy, leading to a 25% decrease in SLA breaches across IS&T and ticket incidents.

Tata 1mg

• Utilized Asyncio, Celery, and RabbitMQ to orchestrate asynchronous task scheduling, improving operational efficiency for 3,000+ employees in mail scheduling and projecting scalability benefits for over 4 million customers.

Cisco

• Conducted SDWAN network data analysis and developed an internal dashboard to visualize Quality of Service (QoS) metrics, achieving a 23.2% improvement in efficiency through the use of Data Science, Python, and networking tools.

Licious

• Migrated major application modules from Swift to Flutter, refactoring 5,000+ lines of code, which resulted in a 30% reduction in man-hours and improved application stability.

YASH Technologies

• Collaborated on the development of an Employee Management System portal using Spring Boot, integrating a database to track attendance for 6,000+ employees, leading to significant cost savings and enhanced HR efficiency.

• Developed CustomNerd/NerdEngine, a configurable framework for domain-specific AI assistants using curated retrieval, prompt engineering, relevance filtering, citation generation, and saved-state deployment.
• Built and maintained specialized assistants including DietNerd, WirelessNerd, InvestorNerd, CloudNerd, NewsNerd, and SpaceNerd across PubMed, 3GPP, finance, cloud, news, and astronomy domains; mentors and contributes to group sub-projects OperationsNerd (vertical-configurable assistants) and HallucinationNerd (citation-permutation verification, with Yuta Nam).
• Added multi-source retrieval and resilient fallback logic for CloudNerd Advanced using Stack Overflow, AWS blogs, Azure blogs, CloudFest feeds, and rate-limit-aware API handling.
• Deployed on AWS EC2/S3 with Cloudflare and AWS Load Balancer, incorporating OTP-based authentication via DynamoDB and AWS SES.

NerdEngine / CloudNerd. Lead author, MDPI Electronics (published 2026). Stack: Python · FastAPI · LangChain · RAGAS · OpenAI · Stack Exchange API.

• Architected a modular RAG framework: 9-stage pipeline (query → cascade retrieval → rerank → synthesis) with config-driven saved-state swap across 7 domain packs.
• Designed CloudNerd cascade mode (P1 strict → P2 title-only → P3 facet → P4 broad keywords → P5 gap-fill); citation recovery lifts dev-set scorable rate 6/20 → 20/20.
• Ran three-way 120-question eval on Stack Overflow: CloudNerd 0.695/0.559/0.693/0.866 vs ODR 0.671/0.862/0.848/0.857 vs AutoGPT 0.578/0.096/0.670/0.823. All 4 winner margins significant (paired permutation, 10k iter). 120/120 completion, 0 failures.

• Built an agentic, explainable job-candidate matching system with candidate, employer, and matchmaking agents for transparent recruitment recommendations.
• Implemented resume/job parsing, semantic retrieval, skill overlap, composite scoring, ranked match generation, score breakdowns, and explanation panels using FastAPI, SentenceTransformers, Chroma/Qdrant, and React.

Status: In preparation. Composite ranking 0.949 nDCG@5 (vs cross-encoder reranker 0.939); 340× faster per query (composite 0.42 ms vs cross-encoder 141.7 ms).

• Designed and shipped cert revocation records for Ownly (NDN-based collaborative PWA): Go WASM codec, TypeScript service bridge, Vue UI; RFC 5280 §5.3.1 TLV wire format, cert-hash canonical names, owner-publisher authorization gate, SVS ALO drop check, keychain negative-cert store.
• Contributed to ndnd, the Go implementation of Named Data Networking — certificate revocation and trust validation primitives.

Result. log-R² = 0.992, MAPE 1.83% on held-out Q9/SF20. Validated across 7 experiments; all 95% CIs exclude zero.

• Pipeline: Rust harness around DataFusion::collect() extracting operator-level elapsed_compute from MetricsSet; Python 3.11 analysis (13 raw metrics → 2 features, log1p, closed-form OLS). 596 per-run CSVs; single-threaded Tokio cut CV from 5–8% to 1–3%.
• Coverage: 5 representative TPC-H queries (Q1, Q3, Q6, Q9, Q18) × 6 scale factors (SF1–SF20) = 30 cells, 300 measured runs, 60 warm-up runs on ~19 GB Snappy Parquet.
• Model: log(wall) = −1.51 + 0.64·log(read) + 0.36·log(proc).
• Validation: 30-fold LOCO MAPE 6.07%, 5-fold LOOQ log-R² 0.992±0.007, 6-fold LOSFO MAPE 5.62%; mixed-effects p<0.0001; sample-weighted fit cut Q9/SF20 residual 49.8%; bootstrap 1,000 samples.

• Active research on a scientific-literature query system with a provider-agnostic LLM layer and retrieval flow for moving across model APIs without binding the system to one provider.
• Refactored the LLM layer across multiple providers and kept retrieval, model behavior, and testing boundaries separate so provider changes are easier to reason about.

Bayer-pattern demosaicing research — implemented bayerize + demosaic pipeline in Python; analysis and reproducible scripts alongside a master research document.

B.E., Computer Science | GPA: 9.71

• Awarded the TIET Merit Scholarship for ranking in the top 0.5% all four years of academic studies.
• Ranked 89th out of 100,000 students in the IICC by AICTE coding competition.
• Featured in Apple's Innovation Deck for an idea on Apple Pay, selected out of 350+ submissions.
• Proficient in subjects such as Data Science, Object-Oriented Programming, Operating Systems, Computer Networks, Computer Architecture, and Database Management Systems.
• Recognized as one of the top coders in my institute.
• Actively participated in hackathons and coding competitions.
• Engaged in extracurricular activities, including cricket tournaments, quiz competitions, and debate competitions.
• Served as a Class Representative and Placement Coordinator for my college.
• Vice Lead Ambassador at Thapar Institute Counselling Cell, dedicated to improving mental health hygiene on campus.

Mount Assisi School

Class XII (ISC) | GPA: 96.25%

• Achieved the top position in the Science stream at school.
• Developed a strong understanding of various subjects, including Physics, Chemistry, Mathematics, Literature, and Computer Science.
• Represented the school in FICE seminars and regional debate competitions, displaying excellent communication and critical thinking skills.
• Recognized as the ISC 2019 School Topper, demonstrating exceptional academic achievements and dedication to excellence.
• Showcased strong leadership skills as a School People Leader, effectively guiding and inspiring fellow students.
• Selected as a representative of the school at numerous seminars and FICE events, highlighting excellent communication and presentation abilities.
• Successfully coordinated and managed annual functions and extracurricular events, displaying effective multitasking skills.

Mount Assisi School

Class X (ICSE) | GPA: 94.33%

• Consistently ranked among the top performers throughout my 12 years of schooling, demonstrating exceptional academic achievements.
• Actively engaged in various competitions, such as quizzes, declamations, dances, and debates, where I showcased my exceptional skills and consistently delivered impressive performances.
• Demonstrated my athletic abilities by participating in track and field events and cricket tournaments, achieving notable successes and contributing to the team's accomplishments.
• Proudly represented my school at prestigious national-level Olympiads and the National Talent Association, showcasing my intellectual capabilities and competing at the highest levels of academic excellence.