Open to opportunities · graduated May 2026 Buffalo, New YorkF-1 OPT · 3 years US work authorization

SW Validation Lead & AI/ML Engineer

Niharika Reddy
Katakam Prashanthi.

6.2 years building and validating safety-critical software at Hyundai MOBIS, delivering Electronic Parking Brake systems to GM, Bosch, Stellantis, and HKMC under ISO 26262 and ASPICE. That same rigor now powers production agentic AI systems.

See the journeyniharikareddykatakamprashanthi@gmail.com
Niharika Reddy

Niharika Reddy

SW Validation Lead & AI/ML Engineer

01

Safety-Critical Software

Derives full V-cycle: SRS · SADS · SUDS · SRTS · SITS · SUTS

02

Hardware-in-the-Loop Validation

MIL · SIL · HIL · Fault Injection

03

Applied AI / ML

RAG · LangGraph · Agentic Pipelines · RAGAS

Shipped to

GMBoschStellantisHKMCAudiGMBoschStellantisHKMCAudi

About

Safety-critical discipline, applied to AI.

Owned the full V-cycle on Electronic Parking Brake projects. That meant requirements and design specs (SRS, SADS, SUDS), test specs (SRTS, SITS, SUTS), software reviews, MIL, SIL, and HIL validation with fault injection, FRACAS corrective action, and DVPR evidence packaging. Held three roles at the same time as Technical Lead, Scrum Master, and Product Owner.

Led the SRS/SRTS work for the Audi PASS along with the corrective-action packages and evidence documentation. The EPB project cleared ASPICE CL1 with zero critical non-conformances, and was set up for CL2 readiness.

Spent 1.5 months in Korea on a MOBIS India EPB business trip, working with the MTCK team to align ASPICE processes ahead of the project pass. Brought that knowledge back to train junior engineers at MTCI.

Now applying that rigor to AI systems, building production agentic RAG pipelines, running retrieval benchmarks, and writing reproducible evaluation harnesses.

6.2yrs

Safety-critical SW engineering

4OEMs

GM · Bosch · Stellantis · HKMC

0NCs

Critical non-conformances in audit

30%

EPB test cycle time reduction

Step 01

Define the failure mode

Behind every requirement is a way it fails. Name it on paper before the field does.

Step 02

Build the test

MIL, SIL, HIL, fault injection. The test is the contract between the requirement and the field.

Step 03

Measure the outcome

FRACAS for brakes. RAGAS for retrievers. Numbers, not narratives.

Step 04

Close the gap

Fix it. Document why. Revalidate. Evidence ships with the patch.

Career journey

From Graduate Trainee to Technical Lead in six years.

Three promotions through Hyundai MOBIS Technical Centre. Now back in graduate school, redirecting the same V-cycle discipline toward agentic AI.

  1. Aug 2024 – May 2026Graduated

    M.S. Computer Science

    University at BuffaloBuffalo, NY

    Graduate research focused on retrieval-augmented generation, agentic systems, and ML evaluation. Built a production 3-agent RAG diagnostic pipeline over a 3,178-page vehicle manual corpus (RAGAS faithfulness 0.82, context relevance 0.95) and benchmarked 6 RAG architectures on Apple's FY2024 10-K, achieving BERTScore F1 of 0.99.

    Tools
    LangChainLangGraphPythonDockerRAGASBERTScoreVector DBs
  2. Apr 2024 – Aug 2024Lead

    Lead Research Engineer

    Hyundai MOBIS · MTCI R&DHyderabad, India

    Technical Lead, Scrum Master, and Product Owner for the EPB project simultaneously, leading MBD engineering in MATLAB and Simulink across the full V-model lifecycle: from requirements development through MIL, SIL, and HIL validation, delivering to GM, Bosch, Stellantis, and HKMC.

    • Stepped into Technical Lead ownership of EPB, guiding the team through MATLAB and Simulink development and ensuring consistent delivery across concurrent OEM projects.
    • Established full V-cycle traceability across SRS, SADS, SUDS, SRTS, SITS, and SUTS, and set audit-ready documentation standards that enabled the team to pass external ASPICE assessments.
    • Executed MIL, SIL, and HIL validation including fault injection and failsafe testing. Diagnosed CANape signal anomalies with cross-functional teams to preserve on-time OEM delivery.
    • Took full ownership of SRS and SRTS documentation for the Audi PASS, and led the corrective-action packages and evidence documentation. The EPB project cleared ASPICE CL1 with zero critical non-conformances, and was set up for CL2 readiness.
    • Built MATLAB M-script automation that cut EPB test cycle time by ~30% across MIL, SIL, and HIL phases.
    • Delivered OEM review packages to GM, Bosch, Stellantis, and HKMC. Diagnosed and resolved test failures by implementing fixes in MATLAB and Simulink models, analyzing root causes, documenting justifications, and revalidating to OEM acceptance criteria.
    • Managed sprint backlog, Agile ceremonies, and project visibility in JIRA and Confluence across OEM workstreams. Mentored junior engineers on EPB architecture, ASPICE, and JIRA, tracking team workload to improve sprint capacity and onboarding.
    Standards
    ASPICEISO 26262V-Model SDLC
    Tools
    MATLABSimulinkCANapePXFaultVectorCastPolyspaceEnterprise ArchitectJIRAPolarionIntegrity
  3. Apr 2022 – Mar 2024

    Senior Research Engineer

    Hyundai MOBIS · MTCI R&DHyderabad, India

    Drove HIL validation, fault injection, and failsafe testing across concurrent OEM projects. Established full V-cycle traceability (SRS, SADS, SUDS, SRTS, SITS, SUTS) and audit-ready documentation standards enabling external ASPICE assessments.

    Standards
    ASPICEISO 26262
    Tools
    MATLABSimulinkCANapeDataminerJIRA
  4. Apr 2019 – Mar 2022

    Research Engineer

    Hyundai MOBIS · MTCI R&DHyderabad, India

    Executed MIL / SIL / HIL validation and diagnosed CANape signal anomalies with cross-functional teams to preserve OEM delivery. Resolved test failures by analyzing root causes, implementing model fixes, and revalidating to OEM acceptance criteria.

    Tools
    MATLABSimulinkCANapeVectorCast
  5. Jul 2018 – Mar 2019

    Graduate Engineering Trainee

    Hyundai MOBIS · MTCI R&DHyderabad, India

    Onboarded into the EPB project and the V-model SDLC. Trained on MBD engineering in MATLAB / Simulink and on the toolchain underpinning safety-critical automotive software delivery.

  6. Jun 2014 – May 2018

    B.E. Computer Science

    Karunya Institute of Technology and SciencesCoimbatore, India

    Bachelor of Engineering, Computer Science.

Stack

The toolchain.

The disciplines, tools, and standards I bring to a project, spanning safety-critical validation and modern AI engineering.

Safety & Process

01 / 4
ISO 26262ASPICE (CL1 / CL2)V-Model SDLCFRS / SRS / SRTSFRACASDVPRMBSEFunctional Safety

Validation & Tooling

02 / 4
HILMILSILFault InjectionMATLABSimulinkCANapePXFaultVectorCastPolyspaceEnterprise Architect

AI / ML

03 / 4
RAGAgentic AILangChainLangGraphOpenAIHuggingFaceFAISSChromaDBPineconeNeo4jBM25HyDERAGASBERTScoreFastAPIStreamlitPythonDocker

Delivery & Collaboration

04 / 4
Scrum MasterProduct OwnerJIRAConfluencePolarionIntegrityOEM Review PackagesMentorship

Portfolio

Recent work.

Two open-source projects from earlier this year. Code is on GitHub for both.

Agentic RAG

Intelligent Vehicle Diagnostics Copilot

Three LangGraph agents over a 3,178-page repair-and-owner manual corpus and an OBD-code database. HyDE retrieval on FAISS, served through a FastAPI backend and Streamlit UI, scored end-to-end with RAGAS.

graph.flow● running
retrieve
→
reason
→
answer
RAGAS faithfulness0.82
LangGraphLangChainOpenAIFAISSFastAPIStreamlit
View on GitHub
Benchmark

RAG FINAnalyser

Six RAG architectures benchmarked against 10 financial questions from Apple's FY2024 10-K, scored with RAGAS and BERTScore.

bench.f1 max4 / 6 ran
Traditional
0.94
Hybrid
0.99
Multimodel
0.99
Vectorless
0.95
BERTScore F1 max0.99
LangChainOpenAIPineconeNeo4jBM25RAGAS
View on GitHub

Contact

Let's build something that ships.

I'm available for full-time vehicle software validation and AI / ML engineering roles, remote or on-site, anywhere in the US. Graduated May 2026, on F-1 OPT. The fastest way to reach me is email.

Email meLinkedInGitHub

Based in

Buffalo, New York

Authorization

F-1 OPT · 3 years US work authorization

Availability

Available now · MS CS, May 2026