Agentic products
Natural-language systems that inspect evidence, propose actions, create editable drafts, and keep humans in control.
Open to full-time software and AI product roles
FULL-STACK ENGINEER / AI PRODUCTS
I build systems that move from insight to action.
I build production AI systems that turn complex data into controlled, useful action.
4+Production AI workflows
100M+Records queried
1669Codeforces peak
750+Problems solved
01 / PROFILEEngineering decisions,
Engineering decisions,
not just demos.
Full-stack developer focused on agentic AI, distributed backends, and customer-facing products. I own systems end to end - from APIs, queues, databases, and model orchestration to the interfaces people use to make decisions.
Full-stack ownership
Features carried from schema and API design through queues, integrations, React workflows, deployment, and production support.
Systems under constraints
Fast, reliable software built around real limits: large datasets, small GPUs, external APIs, and multi-tenant infrastructure.
02 / EXPERIENCEBuilt, shipped,
Built, shipped,
kept running.
Four roles spanning production AI products, applied machine learning, infrastructure, and measurable customer outcomes.
01
Full Stack Developer - AI Products
CausalFunnel ↗Own AI-powered marketing systems across paid growth, SEO/GEO, and conversion optimization from backend architecture through customer-facing workflows.
- Architected a natural-language Google Ads agent using TypeScript, Fastify, PostgreSQL, and Google Ads APIs to audit performance, tracking, search-term waste, budgets, and bidding
- Enabled controlled campaign creation and optimization through editable drafts, schema validation, approvals, audit trails, asynchronous execution, and provider read-back verification
- Built a multi-tenant WordPress SEO automation agent with Fastify, Prisma, PostgreSQL, Redis, and BullMQ, including draft, publish, verification, and rollback workflows
- Extended SEO into generative engine optimization across multiple AI search surfaces, tracking mentions, citations, competitors, and answer quality
- Shipped an agentic CRO workflow using traffic, heatmaps, user journeys, and funnel signals to propose nudges, A/B tests, and landing-page improvements
TypeScriptReactFastifyPostgreSQLMongoDBRedisBullMQGoogle Ads APIWordPress
02
AI Intern
Cynapto Technologies
Worked on audiovisual synchronization and neural rendering pipelines under constrained compute.
- Designed an end-to-end audiovisual synchronization pipeline using bidirectional LSTM networks
- Improved synchronization accuracy by 30% through sequence-window, model, and feature-alignment experiments
- Profiled and optimized NeRF rendering pipelines, reducing processing time by 25% while balancing memory and visual quality
PythonBiLSTMNeRFPyTorchComputer Vision
03
AI Quality Analyst
KrossWork Solutions
Built applied NLP and backend systems for call auditing, classification, and customer segmentation.
- Designed a BERT and rule-based call-auditing system analyzing more than 10,000 calls
- Built a sentiment-aware classifier with 87% accuracy across 1,000+ labeled calls
- Created eight customer segments that increased conversions by 25%
- Reduced REST API latency by 40% through LRU caching and lazy loading
PythonBERTNLPREST APIsCaching
04
Data & ML Intern
Prabha Enterprises
Applied machine learning to delivery forecasting, inventory planning, visual inspection, and supply-chain risk.
- Designed XGBoost regression models that reduced delivery times by 35%
- Built a Random Forest inventory pipeline that reduced costs by 20%
- Deployed ResNet-50 defect inspection with 95% accuracy
- Forecast supply-chain disruptions with LSTM models at 82% precision
PythonTensorFlowXGBoostResNet-50LSTM
03 / SELECTED WORKProof through
Proof through
working systems.
Six projects across large-scale data, constrained inference, explainable AI, knowledge graphs, computer vision, and agentic execution.
Approximate query processing at 100M+ record scale
TrendForage
5xfaster execution
Architected an approximate query-processing engine that avoids full scans by using probabilistic stream sampling with measurable accuracy bounds.
- Queries datasets containing more than 100 million records
- Benchmarks latency, throughput, sample size, and approximation error in real time
- Produced findings that contributed to two co-authored research publications
Domain AI inference on constrained GPU memory
Galenos
3.8Bparameters fine-tuned
Built an end-to-end domain-specific assistant spanning data preparation, model fine-tuning, optimized inference, and real-time streamed responses.
- Fine-tuned Phi-3 with QLoRA and 4-bit quantization
- Fit model inference inside the 15 GB memory limit of an NVIDIA T4
- Engineered Socket.io streaming and efficient context handling
Explainable deep-learning market pattern engine
CognitiCharts
87%pattern accuracy
Combined TensorFlow and PyTorch pattern recognition with SHAP-based explainability to reduce false positives and make model decisions inspectable.
- Reduced false positives by 28% with explainable AI
- Reached a 2.3 Sharpe ratio on five years of backtested data
- Built during Sahamati BuildAAthon 2024
Graph-RAG engine grounded in Ayurvedic knowledge
Atreya
90%grounded accuracy
Created a grounded retrieval system with LangChain and Neo4j Aura over a structured knowledge graph of more than 1,000 herbs.
- Improved retrieval precision by 75%
- Measured a 40% improvement in treatment efficacy
- Built for Smart India Hackathon and Social Summer of Code
Species classification and fishery yield optimization
FinAlogica
95%classification accuracy
Delivered a MobileNetV3 classifier behind FastAPI and connected real-time weather signals to a PostgreSQL-backed yield optimization workflow.
- Used transfer learning for accurate species classification
- Added RBAC and bcrypt-backed access control
- Boosted modeled fishery productivity by 30%
Simulation-first agentic tooling for EVM workflows
BlockAgent Toolkit
40%less deployment time
Orchestrated autonomous contract workflows with Web3.js and REST APIs while using controller-led simulation as a safety boundary.
- Handled more than 1,000 monthly transactions
- Designed controller logic for safer contract execution
- Built across Celo, Aurora, and Harmony ecosystems
04 / ENGINEERING STACKBroad enough to own it.
Broad enough to own it.
Deep enough to debug it.
From DNS and TLS at the edge to queues, model APIs, databases, and the product interface.
Networking & Web Internals
DNSHTTP/HTTPSTLSTCPReverse ProxyingWebSocketsNginxNetwork Debugging
Languages
TypeScriptJavaScriptPythonJavaC++Rust
Full-Stack & APIs
ReactNext.jsNuxt.jsNode.jsExpressFastifyFastAPIFlaskREST APIsWebhooks
Backend & Distributed Systems
MicroservicesEvent-Driven ArchitectureApache KafkaBullMQBackground WorkersCachingApache Spark
Data
PostgreSQLMongoDBMySQLRedisNeo4jPrismaQuery OptimizationIndexing
Authentication & Security
JWTOAuth 2.0Auth0RBACbcryptTenant IsolationRate LimitingSSRF Protection
Cloud & DevOps
AWS EC2AWS S3DockerDocker ComposeKubernetesGitHub ActionsCI/CDLinux/Bash
Integrations
Google Ads APIWordPress APIShopify APIStripeRazorpaySocket.io
AI & LLM Engineering
Multi-provider LLM APIsRAGLangChainQLoRATensorFlowPyTorchPrompt EngineeringTool Calling
Testing & Core CS
PytestJestVitestPostmanIntegration TestingData Structures & AlgorithmsOOPSystem Design
05 / PROBLEM SOLVING
Practice that compounds.
Algorithms sharpen implementation. Systems thinking makes the implementation useful in production.
750+problems solved
1669Codeforces peak
Expertcompetitive rank
06 / RECOGNITIONMeasured beyond
Measured beyond
the project list.
Industry-sponsored Kaggle ML competition
Ranked among 2,700+ submissions
Codeforces competitive programming rank
Peak rating of 1669
Code with Cisco
Recognized among top Software Engineer candidates
GATE Computer Science 2025 & 2026
Qualified twice in the national examination
LeetCode problems solved
Data structures, algorithms, and system design
07 / EDUCATION
JSS Academy of Technical Education, Noida
Bachelor of Technology in Computer Science and Engineering
- Period
- 2022 - 2026
- CGPA
- 8.0 / 10
- Status
- Graduating 2026