Founding Engineer @ DAITA. IIT Patna CS grad.
I build 0→1 products, data systems, and AI pipelines.
Knight on LeetCode (2004). Ship fast. Break things. Fix them faster.
❯ whoamiRajneesh Kumar — Founding Engineer ❯ cat profile.json{"college": "IIT Patna","degree": "B.Tech CSE","cpi": 8.02,"current": "DAITA","leetcode": "Knight (2004)","problems_solved": 1200+} ❯ git log --oneline -4a1f9e12 feat: AI chatbot w/ near-100% accuracy @ Daitak7u6p90 feat: TNA engine w/ near-99% time saved in creating Time & action plan for production @ Daita93cne12 feat: Dynamic Rule Engine w/ instant fraud prevention @ Meeshod8kf903 fix: Migrating older python services to Java @ Meesho ❯
01.
Work Experience
Oct 2025 — Present
Current
Founding Engineer
DAITA
Architected and built multiple core operational systems from 0→1 for a fast-moving manufacturing/workflow platform, including PO document parsing, supplier parsing, TNA generation workflows, supplier chasing systems, notifications, chat infrastructure, and operational automation pipelines.
Designed scalable backend workflows for parsing and transforming unstructured operational documents into structured production workflows, enabling automated onboarding and downstream execution pipelines.
Built and maintained supplier communication infrastructure including WhatsApp-based messaging, automated chasing systems, notification orchestration, morning operational alerts, and real-time workflow coordination features.
Rapidly shipped high-impact product systems under evolving requirements and tight operational timelines, frequently delivering production-ready features within days while maintaining system stability and cross-service compatibility.
Worked closely with engineering leadership on architectural tradeoffs, scalability concerns, and long-term operational system design for manufacturing and supplier workflow management.
Proactively handled production stability, onboarding recovery flows, edge-case debugging, and live operational issue mitigation to reduce downstream user impact during critical incidents.
Continuously refactored and stabilized backend systems to support evolving product requirements, operational scale, and cross-functional workflows without introducing major production regressions.
Collaborated across product, operations, and engineering teams to align technical implementation with real-world operational constraints and manufacturing workflows.
Built systems with a strong focus on scalability, reliability, and operational resilience in a high-speed startup environment with rapidly changing business requirements.
Developed a Dynamic rule engine to execute analytical rules (using MVEL) for fraud prevention in order and returns. Integrated multiple services and data sources via runtime wrappers for efficient rule evaluation.
Migrated multiple legacy Python-based service to Java Spring Boot, added structured debugging logs and Prometheus-based metrics, enhancing fault tolerance and reducing latency.
worked on reducing RTO(return to origin) frauds and blocked orders at cart level which was going to get blocked in later phases of order processing like user-supplier interaction history check, high rto history checks, past delivery fraud checks(wrong product getting returned by user instead of original product) and high rto rate users. Which helped platform to reduce the drag and unnecessary resource burn for orders which was getting cancelled in intermittent stages with in the system.
implimented COD blocking and forced pre-payment for low intent high RTO user which improved over all delivered product acceptance and achieved reduction in over all RTO rate for such user cohorts (idea: highly unlikely that user will not accept a prepaid order on delivery unlike cod where user can reject the order because user haven't paid yet and they got nothing to lose)
Performing continuous infrastructure monitoring and conducting root cause analysis to identify and resolve production issues, improving system stability.
JavaSpring BootMVELPrometheusFraud Prevention
Jun 2024 — Dec 2024
Past
Full Stack Developer
RAAPID INC
Redesigned Entire Production Database structure to align with FHIR-compliant industry standards. Improving existing Events and Procedures.
Build archival machanism for moving older data from active tables to cold storage to improve query performance and lower the cost associated with database.
Streamlined report generation by replacing Python-based workflows with efficient SQL solutions, reducing complexity, enhancing communication, cutting review time, improving accuracy, and minimizing logical errors.
Worked on Azure data pipelines, automated data insertion tasks, and optimizing query performance through indexing, locking, and server parameter adjustments.
Optimized SQL Performance - Implemented table partitioning and preprocessed raw data to significantly enhance query execution time and database efficiency.
LLM-Powered SQL Generation - Developed LLM agents using Phidata with a structured knowledge base of the product database, enabling automatic generation of complex SQL queries through intelligent prompting.
LLM-Driven Code & Test Analysis - Built a multi-agent LLM system for repository analysis, empowering users to navigate complex codebases without prior knowledge. The system also automated test case generation for new features, accelerating the QA process.
SQLAzure Data FactoryNode.jsAngular
Jan 2024 — May 2024
PAST
Software Developer
RAAPID INC
Created default and custom triggers to monitor 1,000+ individual user sessions daily on the product page via Matomo, storing data in MySQL.
Calculated active and idle time per session using RDBMS Events and Procedures, achieving 95%+ accuracy.
Visualized performance metrics (time spent, feature usage, user flow) using Metabase — zero cost.
MatomoMySQLMetabaseAnalytics
May 2023 — Jul 2023
Intern
SDE Intern — DPDK R&D
Truminds Software Systems
Developed efficient C code for high-performance packet processing in DPDK, focusing on GTP packet identification, decapsulation, and advanced filtering (port&IP).
Implemented optimized interrupt handling in DPDK to reduce latency and enhance packet processing and filtering.
Leveraged low-level C programming and hardware acceleration in DPDK for significant performance improvements in packet manipulation.
Demonstrated around 3-4x speedup in raw packet processing over traditional stack utilising higher network bandwidth and overcoming the bottle necks of packet processing of osi model by moving the data link, network & transport layer from kernal to user space eliminating all the context switching of user and kernal space, hardware interrupts & copying of packets from kernal memory to user memory.
Effectively moved entire packet processing from kernal to user space. directly accessing packets from nics and writing directly to user-space memory without involving cpu in copy process(zero copy process).
CDPDKNetworkingLow Latency
02.
Projects
🔭
Hackathon Winner · Meesho Hackme3
Observability Platform
Automated triage for broken microservices—because nobody wants to read thousands of raw log lines during an outage
Built with Java, Python, gRPC, and VictoriaMetrics to collect, store, and query logs/metrics across microservices
Reduced detection time by 80–90% and achieved 85%+ accuracy across services
AI agents for threshold-based anomaly detection and root cause analysis — dynamic on-demand inference to avoid always-on model costs
A multimodal smart assistant with real memory, eyesight, and zero robotic "Sorry, I didn't get that."
Concurrent by design: Built in Go using lightweight goroutines to ingest real-time audio streams, trigger computer vision, and execute tools without UI thread lockup.
Hybrid Edge/Cloud brain: Dynamic model routing that runs local weights for zero-latency instant commands, deferring to cloud LLMs only for heavy-duty reasoning.
Full tool execution: Native integrations for streaming media playback, camera feeds, and automated workflow triggers (email, system tasks).
Dynamic chat platform with private messaging, group formation, video calling, screen sharing, collaborative whiteboard, and file sharing — powered by Firebase + Jitsi.
Voice-controlled Python assistant — search the web, play music, open apps. Achieved 80% speech recognition accuracy using SpeechRecognition and pyttsx3.